diff --git a/reports/benchmark_reproducibility.json b/reports/benchmark_reproducibility.json index 0e32dcd..5e89306 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": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "git_status": { "available": true, "dirty": false, @@ -17,9 +17,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429", - "source_contract_sha256": "52472a5f16d07e88959326df916a03a327b3f3c9443d2128556002c431c9f70a", - "archive_sha256": "36ec77ec0badfe8b44286434d0c10de0470dee241f33b569ea24a0e971721a0e", + "evidence_bundle_sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf", + "source_contract_sha256": "e19dbe7ff57701690de54324ba7f20bca0f7f3715f73f8721fbcc8c33f5b9b25", + "archive_sha256": "9bcb614b9c6eb6f38db0ab71019518f3f2320f060237c732d32bfd2c3e01bf23", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -50,7 +50,7 @@ }, "release_lock": { "ready": true, - "commit": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "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": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429" + "sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -133,8 +133,8 @@ "label": "output_execution", "path": "reports/output_execution_runs.json", "exists": true, - "bytes": 7967, - "sha256": "f934e2b43e24d30cd23098b81c7d2602d9038c793f1afd7cd42b64974bee3bc2" + "bytes": 7966, + "sha256": "2991764f1ab41e6b2c37ddffa9b8bb51d967719487d3b4ade408dd077a182d91" }, { "label": "blind_review", @@ -169,7 +169,7 @@ "path": "reports/security_trust_report.json", "exists": true, "bytes": 99433, - "sha256": "bde78b81ac27913186aa6072f60bddb8fc2b006f596413e032e4864ad345c1d3" + "sha256": "016c810beee5c9020f5862ad0a564b5f6d760d8da481e38de9e36ec4a9a89e77" }, { "label": "python_compatibility", @@ -183,14 +183,14 @@ "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "28c7a971924e56012df29e7bd924cc7a2d9a6de78b37b327b81c3b222cad29dd" + "sha256": "4deadc85347967c1b48b7e0648b4e3e70faa6b9a630fbef2656e039d2d75d26b" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19325, - "sha256": "34d291fa975346752b34283f38f60e3498a7d1165f89f4081200ad8c2a8d5e1b" + "sha256": "55f5816930f21659000f1eaaea9bdf6d600baf4bee6c1d1c4597714bf6a8c464" }, { "label": "install_simulation", @@ -217,8 +217,8 @@ "label": "world_class_evidence_ledger", "path": "reports/world_class_evidence_ledger.json", "exists": true, - "bytes": 11613, - "sha256": "0bff14542475657a1261c1afcd87ffabc5d5c5e75bcd20afcfeacc44b16b7c14" + "bytes": 16763, + "sha256": "7b3d12cd30870cb361b7bb09e43d5ea88fd5bbe35a6d78f91303862a71489ff2" }, { "label": "world_class_evidence_intake", diff --git a/reports/benchmark_reproducibility.md b/reports/benchmark_reproducibility.md index 578179c..3ca6af6 100644 --- a/reports/benchmark_reproducibility.md +++ b/reports/benchmark_reproducibility.md @@ -1,9 +1,9 @@ # Benchmark Reproducibility Generated at: `2026-06-14` -Commit: `0936ce727e7ebc414bcc5a95db139bbecbd594f9` +Commit: `9067c9e211bd0bde80eb0a65b797d01426c97a81` Working tree dirty at generation: `false` -Evidence bundle SHA256: `ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429` +Evidence bundle SHA256: `72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf` ## Summary @@ -12,8 +12,8 @@ Evidence bundle SHA256: `ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197 - methodology complete: `true` - required artifacts: `24` - missing artifacts: `0` -- source contract sha256: `52472a5f16d0` -- archive sha256: `36ec77ec0bad` +- source contract sha256: `e19dbe7ff577` +- archive sha256: `9bcb614b9c6e` - 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: `ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429` +- sha256: `72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf` ## Methodology Sections @@ -70,19 +70,19 @@ 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 | `f934e2b43e24` | +| output_execution | `reports/output_execution_runs.json` | present | `2991764f1ab4` | | 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 | `bde78b81ac27` | +| trust_report | `reports/security_trust_report.json` | present | `016c810beee5` | | python_compatibility | `reports/python_compatibility.json` | present | `ae16e17266e4` | -| registry_audit | `reports/registry_audit.json` | present | `28c7a971924e` | -| package_verification | `reports/package_verification.json` | present | `34d291fa9753` | +| registry_audit | `reports/registry_audit.json` | present | `4deadc853479` | +| package_verification | `reports/package_verification.json` | present | `55f5816930f2` | | install_simulation | `reports/install_simulation.json` | present | `1ed7ce5266c2` | | skill_os2_audit | `reports/skill_os2_audit.json` | present | `6bb2dcb0e1e5` | | world_class_evidence_plan | `reports/world_class_evidence_plan.json` | present | `164803bb1cea` | -| world_class_evidence_ledger | `reports/world_class_evidence_ledger.json` | present | `0bff14542475` | +| world_class_evidence_ledger | `reports/world_class_evidence_ledger.json` | present | `7b3d12cd3087` | | world_class_evidence_intake | `reports/world_class_evidence_intake.json` | present | `8f7f9db6f899` | | world_class_submission_review | `reports/world_class_submission_review.json` | present | `a6d609911ed5` | | world_class_operator_runbook | `reports/world_class_operator_runbook.json` | present | `bb416ed28648` | diff --git a/reports/review-studio.json b/reports/review-studio.json index 4881340..f805afe 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": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429", - "source_contract_sha256": "52472a5f16d07e88959326df916a03a327b3f3c9443d2128556002c431c9f70a", - "archive_sha256": "36ec77ec0badfe8b44286434d0c10de0470dee241f33b569ea24a0e971721a0e", + "evidence_bundle_sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf", + "source_contract_sha256": "e19dbe7ff57701690de54324ba7f20bca0f7f3715f73f8721fbcc8c33f5b9b25", + "archive_sha256": "9bcb614b9c6eb6f38db0ab71019518f3f2320f060237c732d32bfd2c3e01bf23", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -1357,7 +1357,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "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.", @@ -4917,7 +4917,7 @@ "ok": true, "generated_at": "2026-06-14", "skill_dir": ".", - "commit": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "git_status": { "available": true, "dirty": false, @@ -4931,9 +4931,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429", - "source_contract_sha256": "52472a5f16d07e88959326df916a03a327b3f3c9443d2128556002c431c9f70a", - "archive_sha256": "36ec77ec0badfe8b44286434d0c10de0470dee241f33b569ea24a0e971721a0e", + "evidence_bundle_sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf", + "source_contract_sha256": "e19dbe7ff57701690de54324ba7f20bca0f7f3715f73f8721fbcc8c33f5b9b25", + "archive_sha256": "9bcb614b9c6eb6f38db0ab71019518f3f2320f060237c732d32bfd2c3e01bf23", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -4964,7 +4964,7 @@ }, "release_lock": { "ready": true, - "commit": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "status_scope": "generation-time status before this report is written", "reason": "clean generation-time HEAD" }, @@ -4974,7 +4974,7 @@ "existing_count": 24, "missing_count": 0, "missing_paths": [], - "sha256": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429" + "sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -5047,8 +5047,8 @@ "label": "output_execution", "path": "reports/output_execution_runs.json", "exists": true, - "bytes": 7967, - "sha256": "f934e2b43e24d30cd23098b81c7d2602d9038c793f1afd7cd42b64974bee3bc2" + "bytes": 7966, + "sha256": "2991764f1ab41e6b2c37ddffa9b8bb51d967719487d3b4ade408dd077a182d91" }, { "label": "blind_review", @@ -5083,7 +5083,7 @@ "path": "reports/security_trust_report.json", "exists": true, "bytes": 99433, - "sha256": "bde78b81ac27913186aa6072f60bddb8fc2b006f596413e032e4864ad345c1d3" + "sha256": "016c810beee5c9020f5862ad0a564b5f6d760d8da481e38de9e36ec4a9a89e77" }, { "label": "python_compatibility", @@ -5097,14 +5097,14 @@ "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "28c7a971924e56012df29e7bd924cc7a2d9a6de78b37b327b81c3b222cad29dd" + "sha256": "4deadc85347967c1b48b7e0648b4e3e70faa6b9a630fbef2656e039d2d75d26b" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19325, - "sha256": "34d291fa975346752b34283f38f60e3498a7d1165f89f4081200ad8c2a8d5e1b" + "sha256": "55f5816930f21659000f1eaaea9bdf6d600baf4bee6c1d1c4597714bf6a8c464" }, { "label": "install_simulation", @@ -5131,8 +5131,8 @@ "label": "world_class_evidence_ledger", "path": "reports/world_class_evidence_ledger.json", "exists": true, - "bytes": 11613, - "sha256": "0bff14542475657a1261c1afcd87ffabc5d5c5e75bcd20afcfeacc44b16b7c14" + "bytes": 16763, + "sha256": "7b3d12cd30870cb361b7bb09e43d5ea88fd5bbe35a6d78f91303862a71489ff2" }, { "label": "world_class_evidence_intake", diff --git a/reports/review-viewer.json b/reports/review-viewer.json index a95e2d5..5ec4f23 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": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429", - "source_contract_sha256": "52472a5f16d07e88959326df916a03a327b3f3c9443d2128556002c431c9f70a", - "archive_sha256": "36ec77ec0badfe8b44286434d0c10de0470dee241f33b569ea24a0e971721a0e", + "evidence_bundle_sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf", + "source_contract_sha256": "e19dbe7ff57701690de54324ba7f20bca0f7f3715f73f8721fbcc8c33f5b9b25", + "archive_sha256": "9bcb614b9c6eb6f38db0ab71019518f3f2320f060237c732d32bfd2c3e01bf23", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -944,7 +944,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "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 84c1e18..a31fb4a 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": "ba36213f867958bf4a3f35d0d81e864862134510e313b87b4326197b25f36429", - "source_contract_sha256": "52472a5f16d07e88959326df916a03a327b3f3c9443d2128556002c431c9f70a", - "archive_sha256": "36ec77ec0badfe8b44286434d0c10de0470dee241f33b569ea24a0e971721a0e", + "evidence_bundle_sha256": "72e36f4dcd9333a02663a39c209e26770384f247c6a5ad394b4d76684d04bdaf", + "source_contract_sha256": "e19dbe7ff57701690de54324ba7f20bca0f7f3715f73f8721fbcc8c33f5b9b25", + "archive_sha256": "9bcb614b9c6eb6f38db0ab71019518f3f2320f060237c732d32bfd2c3e01bf23", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -943,7 +943,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "0936ce727e7ebc414bcc5a95db139bbecbd594f9", + "commit": "9067c9e211bd0bde80eb0a65b797d01426c97a81", "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.",