diff --git a/reports/benchmark_reproducibility.json b/reports/benchmark_reproducibility.json index 6af364d1..81e62764 100644 --- a/reports/benchmark_reproducibility.json +++ b/reports/benchmark_reproducibility.json @@ -3,44 +3,18 @@ "ok": true, "generated_at": "2026-06-17", "skill_dir": ".", - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "git_status": { "available": true, - "dirty": true, - "changed_file_count": 18, - "generated_dirty": true, - "generated_changed_file_count": 18, + "dirty": false, + "changed_file_count": 0, + "generated_dirty": false, + "generated_changed_file_count": 0, "source_dirty": false, "source_changed_file_count": 0, - "sample": [ - " M reports/adoption_drift_report.json", - " M reports/adoption_drift_report.md", - " M reports/context_budget.json", - " M reports/context_budget_summary.json", - " M reports/output_execution_runs.json", - " M reports/output_execution_runs.md", - " M reports/world_class_evidence_intake.json", - " M reports/world_class_evidence_ledger.json", - " M reports/world_class_evidence_ledger.md", - " M reports/world_class_evidence_plan.json", - " M reports/world_class_evidence_plan.md", - " M reports/world_class_evidence_preflight.json" - ], + "sample": [], "source_sample": [], - "generated_sample": [ - " M reports/adoption_drift_report.json", - " M reports/adoption_drift_report.md", - " M reports/context_budget.json", - " M reports/context_budget_summary.json", - " M reports/output_execution_runs.json", - " M reports/output_execution_runs.md", - " M reports/world_class_evidence_intake.json", - " M reports/world_class_evidence_ledger.json", - " M reports/world_class_evidence_ledger.md", - " M reports/world_class_evidence_plan.json", - " M reports/world_class_evidence_plan.md", - " M reports/world_class_evidence_preflight.json" - ], + "generated_sample": [], "generated_dirty_prefixes": [ "dist/", "registry/index.json", @@ -57,7 +31,7 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403", + "evidence_bundle_sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a", "source_contract_sha256": "5533575c144fb56486ab7320bfbd9f6655a6360cc1c2fbab837c40a7bc57ca6d", "archive_sha256": "7db44f059844b71930ea8194a72ba1a316116ca970620de5550d60616e8f23c2", "output_case_count": 5, @@ -78,12 +52,12 @@ "world_class_source_blocked_count": 10, "public_claim_ready": false, "public_claim_blocker_count": 4, - "working_tree_dirty": true, - "changed_file_count": 18, + "working_tree_dirty": false, + "changed_file_count": 0, "source_tree_dirty": false, "source_changed_file_count": 0, - "generated_tree_dirty": true, - "generated_changed_file_count": 18 + "generated_tree_dirty": false, + "generated_changed_file_count": 0 }, "public_claim": { "ready": false, @@ -98,11 +72,11 @@ }, "release_lock": { "ready": true, - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "status_scope": "generation-time status before this report is written", "source_changed_file_count": 0, - "generated_changed_file_count": 18, - "reason": "only generated evidence artifacts were dirty at generation time" + "generated_changed_file_count": 0, + "reason": "clean source tree at generation-time HEAD" }, "evidence_bundle": { "algorithm": "sha256(path,label,exists,artifact_sha256)", @@ -110,7 +84,7 @@ "existing_count": 25, "missing_count": 0, "missing_paths": [], - "sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403" + "sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -281,8 +255,8 @@ "label": "world_class_evidence_preflight", "path": "reports/world_class_evidence_preflight.json", "exists": true, - "bytes": 97462, - "sha256": "74c0cb843c38aa63434a1a9109e7bccccc60e675b8025183dd8a1b66a9def7b4" + "bytes": 100952, + "sha256": "7bbbf51cfbaffbe606566c1eebfd8b3927e75b33efeea5132e1bd03382c897f9" }, { "label": "world_class_submission_review", diff --git a/reports/benchmark_reproducibility.md b/reports/benchmark_reproducibility.md index 8aa94704..12211f2e 100644 --- a/reports/benchmark_reproducibility.md +++ b/reports/benchmark_reproducibility.md @@ -1,11 +1,11 @@ # Benchmark Reproducibility Generated at: `2026-06-17` -Commit: `28c48ec248807020a68b50e012d4419edafbb418` -Working tree dirty at generation: `true` +Commit: `136654cda911836efd177294a17efc4e5bf08ae9` +Working tree dirty at generation: `false` Source tree dirty at generation: `false` -Generated evidence dirty at generation: `true` -Evidence bundle SHA256: `2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403` +Generated evidence dirty at generation: `false` +Evidence bundle SHA256: `9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a` ## Summary @@ -25,9 +25,9 @@ Evidence bundle SHA256: `2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6a - world-class source checks: `9` pass / `19` total; `10` blocked - public claim ready: `false` - public claim blockers: `4` -- changed files at generation: `18` +- changed files at generation: `0` - source changed files at generation: `0` -- generated changed files at generation: `18` +- generated changed files at generation: `0` This report proves local benchmark reproducibility only. It keeps external provider and human-review gaps visible instead of counting them as complete. The git commit and dirty samples are generation-time context; the evidence bundle SHA is the durable anchor for the artifacts listed below. @@ -47,14 +47,14 @@ This report proves local benchmark reproducibility only. It keeps external provi ## Release Lock - ready: `true` -- reason: only generated evidence artifacts were dirty at generation time +- reason: clean source tree at generation-time HEAD - status scope: generation-time status before this report is written ## Evidence Bundle - algorithm: `sha256(path,label,exists,artifact_sha256)` - artifacts: `25` / `25` -- sha256: `2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403` +- sha256: `9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a` ## Methodology Sections @@ -90,7 +90,7 @@ This report proves local benchmark reproducibility only. It keeps external provi | world_class_evidence_plan | `reports/world_class_evidence_plan.json` | present | `2beb888d7992` | | world_class_evidence_ledger | `reports/world_class_evidence_ledger.json` | present | `22c85613d7be` | | world_class_evidence_intake | `reports/world_class_evidence_intake.json` | present | `d3f44570e605` | -| world_class_evidence_preflight | `reports/world_class_evidence_preflight.json` | present | `74c0cb843c38` | +| world_class_evidence_preflight | `reports/world_class_evidence_preflight.json` | present | `7bbbf51cfbaf` | | world_class_submission_review | `reports/world_class_submission_review.json` | present | `96624a5b7f79` | | world_class_operator_runbook | `reports/world_class_operator_runbook.json` | present | `a7b7809d3b42` | | world_class_operator_runbook_markdown | `reports/world_class_operator_runbook.md` | present | `d72777b421b5` | diff --git a/reports/evidence_consistency.json b/reports/evidence_consistency.json index 1644d621..4a155d2d 100644 --- a/reports/evidence_consistency.json +++ b/reports/evidence_consistency.json @@ -222,7 +222,7 @@ "actual": { "available": true, "clean": false, - "changed_file_count": 28 + "changed_file_count": 6 }, "paths": [ "reports/benchmark_reproducibility.json" @@ -263,8 +263,8 @@ "key": "overview-benchmark-commit", "label": "overview embeds the benchmark commit", "status": "pass", - "expected": "28c48ec248807020a68b50e012d4419edafbb418", - "actual": "28c48ec248807020a68b50e012d4419edafbb418", + "expected": "136654cda911836efd177294a17efc4e5bf08ae9", + "actual": "136654cda911836efd177294a17efc4e5bf08ae9", "paths": [ "reports/benchmark_reproducibility.json", "reports/skill-overview.json" @@ -409,8 +409,8 @@ "key": "interpretation-benchmark-commit", "label": "interpretation embeds the benchmark commit", "status": "pass", - "expected": "28c48ec248807020a68b50e012d4419edafbb418", - "actual": "28c48ec248807020a68b50e012d4419edafbb418", + "expected": "136654cda911836efd177294a17efc4e5bf08ae9", + "actual": "136654cda911836efd177294a17efc4e5bf08ae9", "paths": [ "reports/benchmark_reproducibility.json", "reports/skill-interpretation.json" diff --git a/reports/review-studio.json b/reports/review-studio.json index 299f0cde..48a04bfc 100644 --- a/reports/review-studio.json +++ b/reports/review-studio.json @@ -2389,7 +2389,7 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403", + "evidence_bundle_sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a", "source_contract_sha256": "5533575c144fb56486ab7320bfbd9f6655a6360cc1c2fbab837c40a7bc57ca6d", "archive_sha256": "7db44f059844b71930ea8194a72ba1a316116ca970620de5550d60616e8f23c2", "output_case_count": 5, @@ -2410,14 +2410,14 @@ "world_class_source_blocked_count": 10, "public_claim_ready": false, "public_claim_blocker_count": 4, - "working_tree_dirty": true, - "changed_file_count": 18, + "working_tree_dirty": false, + "changed_file_count": 0, "source_tree_dirty": false, "source_changed_file_count": 0, - "generated_tree_dirty": true, - "generated_changed_file_count": 18 + "generated_tree_dirty": false, + "generated_changed_file_count": 0 }, - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "missing_artifacts": [], "limitations": [ "The git commit and dirty flags are generation-time context; release lock is blocked by source changes, while generated evidence artifacts are tracked separately.", @@ -6219,44 +6219,18 @@ "ok": true, "generated_at": "2026-06-17", "skill_dir": ".", - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "git_status": { "available": true, - "dirty": true, - "changed_file_count": 18, - "generated_dirty": true, - "generated_changed_file_count": 18, + "dirty": false, + "changed_file_count": 0, + "generated_dirty": false, + "generated_changed_file_count": 0, "source_dirty": false, "source_changed_file_count": 0, - "sample": [ - " M reports/adoption_drift_report.json", - " M reports/adoption_drift_report.md", - " M reports/context_budget.json", - " M reports/context_budget_summary.json", - " M reports/output_execution_runs.json", - " M reports/output_execution_runs.md", - " M reports/world_class_evidence_intake.json", - " M reports/world_class_evidence_ledger.json", - " M reports/world_class_evidence_ledger.md", - " M reports/world_class_evidence_plan.json", - " M reports/world_class_evidence_plan.md", - " M reports/world_class_evidence_preflight.json" - ], + "sample": [], "source_sample": [], - "generated_sample": [ - " M reports/adoption_drift_report.json", - " M reports/adoption_drift_report.md", - " M reports/context_budget.json", - " M reports/context_budget_summary.json", - " M reports/output_execution_runs.json", - " M reports/output_execution_runs.md", - " M reports/world_class_evidence_intake.json", - " M reports/world_class_evidence_ledger.json", - " M reports/world_class_evidence_ledger.md", - " M reports/world_class_evidence_plan.json", - " M reports/world_class_evidence_plan.md", - " M reports/world_class_evidence_preflight.json" - ], + "generated_sample": [], "generated_dirty_prefixes": [ "dist/", "registry/index.json", @@ -6273,7 +6247,7 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403", + "evidence_bundle_sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a", "source_contract_sha256": "5533575c144fb56486ab7320bfbd9f6655a6360cc1c2fbab837c40a7bc57ca6d", "archive_sha256": "7db44f059844b71930ea8194a72ba1a316116ca970620de5550d60616e8f23c2", "output_case_count": 5, @@ -6294,12 +6268,12 @@ "world_class_source_blocked_count": 10, "public_claim_ready": false, "public_claim_blocker_count": 4, - "working_tree_dirty": true, - "changed_file_count": 18, + "working_tree_dirty": false, + "changed_file_count": 0, "source_tree_dirty": false, "source_changed_file_count": 0, - "generated_tree_dirty": true, - "generated_changed_file_count": 18 + "generated_tree_dirty": false, + "generated_changed_file_count": 0 }, "public_claim": { "ready": false, @@ -6314,11 +6288,11 @@ }, "release_lock": { "ready": true, - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "status_scope": "generation-time status before this report is written", "source_changed_file_count": 0, - "generated_changed_file_count": 18, - "reason": "only generated evidence artifacts were dirty at generation time" + "generated_changed_file_count": 0, + "reason": "clean source tree at generation-time HEAD" }, "evidence_bundle": { "algorithm": "sha256(path,label,exists,artifact_sha256)", @@ -6326,7 +6300,7 @@ "existing_count": 25, "missing_count": 0, "missing_paths": [], - "sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403" + "sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -6497,8 +6471,8 @@ "label": "world_class_evidence_preflight", "path": "reports/world_class_evidence_preflight.json", "exists": true, - "bytes": 97462, - "sha256": "74c0cb843c38aa63434a1a9109e7bccccc60e675b8025183dd8a1b66a9def7b4" + "bytes": 100952, + "sha256": "7bbbf51cfbaffbe606566c1eebfd8b3927e75b33efeea5132e1bd03382c897f9" }, { "label": "world_class_submission_review", diff --git a/reports/review-viewer.json b/reports/review-viewer.json index 6b2a8d71..2d3d9f83 100644 --- a/reports/review-viewer.json +++ b/reports/review-viewer.json @@ -1010,7 +1010,7 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403", + "evidence_bundle_sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a", "source_contract_sha256": "5533575c144fb56486ab7320bfbd9f6655a6360cc1c2fbab837c40a7bc57ca6d", "archive_sha256": "7db44f059844b71930ea8194a72ba1a316116ca970620de5550d60616e8f23c2", "output_case_count": 5, @@ -1031,14 +1031,14 @@ "world_class_source_blocked_count": 10, "public_claim_ready": false, "public_claim_blocker_count": 4, - "working_tree_dirty": true, - "changed_file_count": 18, + "working_tree_dirty": false, + "changed_file_count": 0, "source_tree_dirty": false, "source_changed_file_count": 0, - "generated_tree_dirty": true, - "generated_changed_file_count": 18 + "generated_tree_dirty": false, + "generated_changed_file_count": 0 }, - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "missing_artifacts": [], "limitations": [ "The git commit and dirty flags are generation-time context; release lock is blocked by source changes, while generated evidence artifacts are tracked separately.", diff --git a/reports/skill-interpretation.json b/reports/skill-interpretation.json index 368c318c..1ffcf537 100644 --- a/reports/skill-interpretation.json +++ b/reports/skill-interpretation.json @@ -1014,7 +1014,7 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403", + "evidence_bundle_sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a", "source_contract_sha256": "5533575c144fb56486ab7320bfbd9f6655a6360cc1c2fbab837c40a7bc57ca6d", "archive_sha256": "7db44f059844b71930ea8194a72ba1a316116ca970620de5550d60616e8f23c2", "output_case_count": 5, @@ -1035,14 +1035,14 @@ "world_class_source_blocked_count": 10, "public_claim_ready": false, "public_claim_blocker_count": 4, - "working_tree_dirty": true, - "changed_file_count": 18, + "working_tree_dirty": false, + "changed_file_count": 0, "source_tree_dirty": false, "source_changed_file_count": 0, - "generated_tree_dirty": true, - "generated_changed_file_count": 18 + "generated_tree_dirty": false, + "generated_changed_file_count": 0 }, - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "missing_artifacts": [], "limitations": [ "The git commit and dirty flags are generation-time context; release lock is blocked by source changes, while generated evidence artifacts are tracked separately.", diff --git a/reports/skill-overview.json b/reports/skill-overview.json index 4af453c6..f43c666b 100644 --- a/reports/skill-overview.json +++ b/reports/skill-overview.json @@ -1009,7 +1009,7 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "2c4ae0e9054c4680f9239be6ec01010ee3f4e72d858c20823a45f6abdd969403", + "evidence_bundle_sha256": "9080619d1499c827ab337b9172cc889106cf64f8c5a8baa86badcc8fc42ecf1a", "source_contract_sha256": "5533575c144fb56486ab7320bfbd9f6655a6360cc1c2fbab837c40a7bc57ca6d", "archive_sha256": "7db44f059844b71930ea8194a72ba1a316116ca970620de5550d60616e8f23c2", "output_case_count": 5, @@ -1030,14 +1030,14 @@ "world_class_source_blocked_count": 10, "public_claim_ready": false, "public_claim_blocker_count": 4, - "working_tree_dirty": true, - "changed_file_count": 18, + "working_tree_dirty": false, + "changed_file_count": 0, "source_tree_dirty": false, "source_changed_file_count": 0, - "generated_tree_dirty": true, - "generated_changed_file_count": 18 + "generated_tree_dirty": false, + "generated_changed_file_count": 0 }, - "commit": "28c48ec248807020a68b50e012d4419edafbb418", + "commit": "136654cda911836efd177294a17efc4e5bf08ae9", "missing_artifacts": [], "limitations": [ "The git commit and dirty flags are generation-time context; release lock is blocked by source changes, while generated evidence artifacts are tracked separately.",