Add bounded manifest and isolation checks for Layer A

cbz/contributions/math514/layer_a/BUILD_REPORT.json

@@ -0,0 +1,146 @@
+{
+  "schema_version": 4,
+  "actual_repository_G1": "NOT_RUN",
+  "current_composed_receipt": "evidence/final-review-identity/current-composed-admission.json",
+  "current_method_inventory": 110,
+  "method_inventory_parts": {
+    "auditor_predecessor": 79,
+    "auditor_review_notebook_new": 9,
+    "harness_predecessor": 15,
+    "harness_identity_new": 7
+  },
+  "current_mutation_definitions": 68,
+  "review_boundary_methods_per_mode": {
+    "cumulative_distinct": 19,
+    "new_total": 9,
+    "affected_predecessor": 10,
+    "latest_notebook_new": 1,
+    "prior_applicable_results_reused": 18
+  },
+  "review_boundary_scanner_mutants_per_mode": {
+    "new_total": 8,
+    "adapted": 3,
+    "cumulative_killed_definitions": 11,
+    "latest_notebook_new": 1,
+    "prior_applicable_kills_reused": 10,
+    "unchanged_definitions_reused": 57
+  },
+  "identity_harness_methods_per_mode": {
+    "normal": 19,
+    "optimized": 19,
+    "new": 7,
+    "affected_predecessor": 12,
+    "unaffected_predecessor_reused": 3
+  },
+  "numeric_metadata_methods_reused": 52,
+  "identity_safeguard_counts": {
+    "themes_per_tool_per_mode": 14,
+    "observations_this_tool": 28,
+    "observations_both_tools": 56,
+    "stages_both_tools": [
+      32,
+      8,
+      4,
+      12
+    ],
+    "single_final_source_campaign": false,
+    "production_scanner_mutants": false
+  },
+  "new_full_G0_suite_or_CI_run": false,
+  "new_Lean_compiler_runs": 0,
+  "new_mathematical_checks": 0,
+  "current_tool_admission": "PASS_SCOPED_COMPOSED_SOURCE_BOUND_COMPONENT_EVIDENCE; actual repository G1 NOT_RUN",
+  "predecessor_evidence": {
+    "schema_version": 3,
+    "actual_repository_G1": "NOT_RUN",
+    "current_composed_receipt": "evidence/declaration-numeric/current-composed-admission.json",
+    "current_method_inventory": 94,
+    "current_mutation_definitions": 60,
+    "fresh_current_repair_methods_per_mode": {
+      "normal": 52,
+      "optimized": 52
+    },
+    "fresh_new_methods_per_mode": 7,
+    "affected_predecessor_methods_per_mode": 45,
+    "reused_predecessor_methods": 42,
+    "new_source_guard_mutants_per_mode": 3,
+    "reused_predecessor_mutations": 57,
+    "new_full_G0_suite_or_CI_run": false,
+    "new_Lean_compiler_runs": 0,
+    "new_mathematical_checks": 0,
+    "predecessor_evidence": {
+      "actual_repository_G1": "NOT_RUN",
+      "all_lazy_profile_generation_guard_mode_executions": 16,
+      "current_composed_receipt": "evidence/lazy-body-profile/current-composed-admission.json",
+      "current_method_inventory": 87,
+      "current_mutation_definitions": 57,
+      "current_repair_checker_mutant_executions": 0,
+      "current_tool_admission": "Composed finite source-bound evidence; lazy-body changed-hunk independent review clear",
+      "fresh_current_repair_methods_per_mode": {
+        "normal": 4,
+        "optimized": 4
+      },
+      "fresh_final_none_profile_grader_guard_mode_executions": 6,
+      "fresh_final_none_profile_grader_guard_mutations": 3,
+      "guard_counts_are_distinct_production_mutants": false,
+      "independent_review": "evidence/lazy-body-profile/shared/independent-lazy-body-review.json",
+      "independent_review_sha256": "8e6c0de322cb2296d2ac9bf472bcb29c3fcdfaa941541539e1fc467d5a47c933",
+      "initial_lazy_type_generation_guard_mode_executions": 10,
+      "method_inventory_parts": {
+        "auditor": 72,
+        "new_lazy_body_profile": 4,
+        "retained_harness": 11
+      },
+      "new_Lean_compiler_runs": 0,
+      "new_full_G0_suite_or_CI_run": false,
+      "new_mathematical_checks": 0,
+      "packaging_execution_verification_runs": 0,
+      "policy": "ordinary synchronous selected methods, every wraps-linked layer; exactly None normal return; no returned-object advancement",
+      "predecessor_evidence": {
+        "actual_repository_G1": "NOT_RUN",
+        "current_composed_receipt": "evidence/composed-admission.json",
+        "current_lexer_component_receipt": "evidence/lexer698-admission.json",
+        "current_method_inventory": 83,
+        "current_mutation_definitions": 57,
+        "current_tool_admission": "Composed finite scoped evidence; current lexer and lifecycle source reviews clear for their stated scopes",
+        "fresh_distinct_methods": 56,
+        "fresh_existing_harness_methods_per_mode": 3,
+        "fresh_mutants_per_mode": 19,
+        "fresh_new_lifecycle_methods_per_mode": 8,
+        "fresh_tests_per_mode": 45,
+        "historical_build_report": "evidence/pre-canonical698-publication/BUILD_REPORT.json",
+        "lexer_component_status": "PASS_AFFECTED_SCOPE",
+        "new_Lean_compiler_runs": 0,
+        "reused_historical_mutants": 38,
+        "reused_unchanged_auditor_methods": 27,
+        "schema_version": 1
+      },
+      "predecessor_metadata": "evidence/pre-lazy-body-profile-publication/BUILD_REPORT.json",
+      "reused_predecessor_methods": 83,
+      "reused_unchanged_initial_lazy_profile_guard_mode_executions": 6,
+      "reused_unchanged_initial_lazy_profile_guard_themes": 3,
+      "schema_version": 2
+    },
+    "predecessor_metadata": "evidence/declaration-numeric/predecessor-metadata/BUILD_REPORT.json",
+    "current_tool_admission": "PASS_AFFECTED_SCOPE; independent exact-source review CLEAR",
+    "independent_review": "evidence/declaration-numeric/review/final-review.json",
+    "independent_review_sha256": "d41632dcd0e600fef55a2065b2a4bfe0ac3905c6f49414b9b960d47073d1d863"
+  },
+  "predecessor_metadata": "evidence/pre-review-identity-metadata/BUILD_REPORT.json",
+  "review_boundary_source_review": "evidence/review-boundary-fix/review/SOURCE_REVIEW.json",
+  "review_boundary_source_review_sha256": "6c305b4cf649d7f371d19d65a8230a403e9bf730fa4043eed2418ac4eb5e8081",
+  "identity_review_status": "CLEAR_SCOPED_ENGINEERING",
+  "current_source_sha256": {
+    "auditor.py": "309935e125195ae8093ffeb691f9d4c16f6c69e3dbb74b453ed6d50e05be1619",
+    "source_checks.py": "6d68a5cb61e41823e0919f13418935f0baef25f41a0d36fb0ec06457e45f4ae0",
+    "mutation_test.py": "9146cd183a8fcdc93623de20596966b118f3a9e5403d94f74f6be0fe46e4e731",
+    "tests/test_auditor.py": "73e0b36fe95774af04b433f2de748911625a4ed10cf9f5f46c86249f19707ddc",
+    "tests/test_mutation_runner.py": "9ecf6dbdc0cf30ecf1b9c4f72d6cda21bc8bfe4eac3a5383ebe4e8b1e93a2866",
+    "canonical_lean.py": "9ed661b1490f73c4dc3664a33ed688e937c1c7d8fecfb4d48c5349fa41823818"
+  },
+  "source_binding_status": "FINAL_ROOT_COMPOSITION_BOUND",
+  "identity_review": "evidence/grader-identity-fix/shared/independent-review-final.json",
+  "identity_review_sha256": "f47d13cc789829966ebf60e7b5e5a9e45059e5ed5db4a5089a3b2ca7534993b4",
+  "notebook_source_review": "evidence/review-boundary-fix/review/NOTEBOOK_SOURCE_REVIEW.json",
+  "notebook_source_review_sha256": "c2edfdb8825fe655a99963ee9fd387ce1bb34c3f85c3989755ec58b5720759d4"
+}

cbz/contributions/math514/layer_a/BUILD_REPORT.md

@@ -0,0 +1,27 @@
+# Layer A composed build report — combined review/identity metadata
+
+## Current combined generation
+
+The combined population is 110 methods: 88 auditor methods (79 numeric-predecessor + 9 review/notebook methods) and 22 harness methods (15 predecessor + 7 new identity methods). The [review-boundary repair](REVIEW_BOUNDARY_FIX.md) cumulatively covers 19 distinct affected methods per mode: 18 applicable prior results plus the one newly checked notebook method. Its 11 affected scanner mutation definitions per mode comprise 10 applicable prior kills plus one newly checked notebook guard; 3 of those 11 are adapted legacy definitions. The 68 active scanner definitions comprise 57 unchanged, 3 adapted and 8 new definitions. The [grader identity repair](GRADER_IDENTITY_FIX.md) covers 19 distinct affected harness methods per mode by composed staged receipts and targeted refreshes (12 predecessor + 7 new); 3 unaffected predecessor harness methods retain their evidence. Identity safeguards are separate grader controls, not scanner mutations.
+
+The completed 52-method numeric-metadata verification remains reusable: Lean/numeric functions are unchanged, and its standard declaration population uses the supported default, no-hook loading profile. Original 18-method review results, the later one-method notebook controls, and identity refreshes remain separately bound. This is not a fresh whole-110-method, whole-68-scanner-mutant, CI, Lean or mathematical run. Current final source bindings and Root's source-partition record are in `evidence/final-review-identity/current-composed-admission.json` and `SOURCE_COMPOSITION.json` beside it. Source-bound original-review, notebook and identity reviews are CLEAR for their declared scopes; they do not authenticate the final repository inventory or scientific statements. Actual repository G1 remains NOT_RUN. Kernel/build provenance, approved axioms, pins and federation admission remain independent. Prior exact-source metadata is retained under `evidence/pre-review-identity-metadata/`.
+
+## Selected-declaration identity profile
+
+The [grader identity repair](GRADER_IDENTITY_FIX.md) freezes actual class, ordinary function and code objects before execution, including wraps-linked functions, and checks them before fixtures, at test starts, at body entry and at body completion. Module coverage comes from the actual module dictionary. Module `load_tests` hooks, callable suite factories, custom constructors/allocation/metaclasses/attribute dispatch and test-method descriptors are refused before construction or invocation. Only exact builtin staticmethod/classmethod descriptors are supported; named selectors traverse ordinary module dictionaries and static class declarations, refusing proxy/descriptor factories before lookup. Ordinary module, class and method selectors remain supported, with the existing synchronous None-return and exact assertion-provenance contracts.
+
+Trusted Python imports still execute. This bounded trusted-runner/source-custody profile is not a sandbox against arbitrary Python or authentication of fabricated consistent JSON. The final independent identity review is CLEAR for its exact source scope. Its 14 safeguard themes per tool in each mode yield 56 generation-bound assertion-kill observations across both tools: 32 + 8 + 4 + 12 across four source generations, not a fresh final-source 56-case campaign or production scanner/replay mutations. Earlier applicable observations and targeted refreshes remain separately bound. See `evidence/grader-identity-fix/CURRENT_COMPOSED_ADMISSION.json`, its shared optimization/guard provenance and `shared/independent-review-final.json` (SHA256 f47d13cc789829966ebf60e7b5e5a9e45059e5ed5db4a5089a3b2ca7534993b4).
+
+## Preserved predecessor record
+
+The numeric-metadata predecessor composition was 94 methods (79 auditor + 15 harness), with 52 affected methods checked in both modes and 42 unaffected methods reused. Its 60 scanner definitions comprised 57 unchanged and 3 new guard mutants checked in both modes. This retained generation is `evidence/declaration-numeric/current-composed-admission.json`; see `NUMERIC_DECLARATION_FIX.md`. It was not a fresh whole-94/60 run and is not the current combined population.
+
+Mutation grading supports ordinary synchronous Python unittest functions or bound methods returning exactly None. Generator, coroutine, async-generator and every wraps-linked lazy method layer are refused during pristine discovery before fixtures or bodies execute. Every non-None return is ERROR, including False, zero, empty strings and opaque iterables. Returned objects are never evaluated for truth, compared for equality, iterated or awaited. Only a fresh native coroutine in CORO_CREATED state may be closed without entering its body. Creating a suspended call object is not evidence that its deferred body executed.
+
+The predecessor lazy-body repair has four distinct new methods in each mode, eighteen fixture receipts per mode with zero lazy-execution sentinels, and three final None-profile safeguard mutations killed by assertions in both modes (six mode executions). The superseded lazy-type generation contains five safeguard themes in two modes (ten mode executions); three unchanged shape, wraps-chain and native-coroutine-disposal themes remain applicable and are reused. These sixteen generation-bound observations are not sixteen distinct production mutants. Packaging ran none of these checks again.
+
+The predecessor receipt remains unchanged at `evidence/composed-admission.json`: its 56 distinct affected methods (45 lexer, 8 lifecycle and 3 retained harness methods) plus 27 unaffected auditor methods cover its 83-method population under their recorded assumptions. Its 19/38 scanner-mutant composition, three one-method dynamic source controls and 10+3+2 generation-bound grader observations remain historical; repeated duplicate-population observations are not distinct production mutants. The original 70/53 and earlier failed generations also remain historical.
+
+Predecessor lazy-body composition: `evidence/lazy-body-profile/current-composed-admission.json`. Its evidence is under `evidence/lazy-body-profile/`: `REPORT.md`, `frozen-manifest.json`, `new-guard-source-controls.json`, normal/optimized logs and fixture receipts. Scoped independent review: `evidence/lazy-body-profile/shared/independent-lazy-body-review.json` (SHA256 8e6c0de322cb2296d2ac9bf472bcb29c3fcdfaa941541539e1fc467d5a47c933); its layer_a section binds the final two runtime hashes. Static original-method shape applicability is preserved alongside it. This review and mathematical evidence do not satisfy an unrelated provider-distinct process requirement.
+
+No fresh whole-suite/G0/CI/build result is created by this package. The underlying auditor, canonical lexer, 72 predecessor behavioral methods, 15 grader methods and 57 predecessor mutation definitions are unchanged; the metadata lexer, seven new tests and three new mutation definitions are the affected runtime scope. Old lifecycle controls and body-trace receipts remain reusable under the static ordinary-None-return shape applicability assessment. Actual repository G1 remains NOT_RUN; final admission is root-owned.