From f26ee325f9b7b25688f062d96b89ac6e280af033 Mon Sep 17 00:00:00 2001 From: Agent IX Date: Sun, 15 Nov 2026 17:34:59 -0800 Subject: [PATCH 1/7] PLAT-1532 extract trust boundary trust level --- spec/functional/FR-004-role-schemas.md | 6 +- .../FR-024-trust-boundary-trust-level.md | 50 +++++++++++ spec/functional/index.md | 1 + spec/tests.md | 9 +- spec_objects_security/manifest.yaml | 4 + .../schemas/TrustBoundary.json | 6 -- .../skeletons/trust_boundary.md | 3 +- tests/test_role_schemas.py | 16 ++-- tests/test_trust_boundary.py | 86 +++++++++++++++++++ typespec/main.tsp | 1 - 10 files changed, 162 insertions(+), 20 deletions(-) create mode 100644 spec/functional/FR-024-trust-boundary-trust-level.md create mode 100644 tests/test_trust_boundary.py diff --git a/spec/functional/FR-004-role-schemas.md b/spec/functional/FR-004-role-schemas.md index a039109..8ce6433 100644 --- a/spec/functional/FR-004-role-schemas.md +++ b/spec/functional/FR-004-role-schemas.md @@ -67,7 +67,7 @@ semantic-core release that renders `identity: false` is a breaking change to the | soup_record | `SoupRecord` | `fields` | `purl: string`, `name: string`, `version: string`, `supplier: string`, `license: string`, `license_fact_ref: SemanticId`, `finding_refs: SemanticId[]`, `inventory_ref: SemanticId`, `serves: SemanticId[]`, `known_anomalies: Anomaly[]`, `support_status: SoupSupportStatus`, `end_of_support: string`, `disposition_clause: string` | `fields` has ≥ 1 item, ≥ 1 identity field, ≥ 1 `purl` field, and 0 defaulted fields; `operations`, `relations` and `clauses` forbidden; no key beyond those listed (the field set of [FR-009](./FR-009-soup-record-identity-and-promotion.md) and [FR-010](./FR-010-soup-record-fields.md)) | | data_classification | `DataClassification` | `fields` | `clauses`, `level: ConfidentialityLevel`, `values: EnumValue[]` | `fields` has ≥ 1 item, ≥ 1 identity field, and ≥ 1 `level` field; `operations` and `relations` forbidden | | attack_surface | `AttackSurface` | `fields` | `clauses`, `relations`, `exposes: SemanticId[]` | `fields` has ≥ 1 item and ≥ 1 identity field; `operations` forbidden | -| trust_boundary | `TrustBoundary` | `fields`, `clauses` | `relations`, `trust_level: TrustLevel`, `crosses: SemanticId[]` | `fields` has ≥ 1 item, ≥ 1 identity field and ≥ 1 `trust_level` field; `clauses` has ≥ 1 item — a boundary that admits every crossing is not a boundary; `operations` forbidden | +| trust_boundary | `TrustBoundary` | `fields`, `clauses` | `relations`, `trust_level: TrustLevel` | `fields` has ≥ 1 item, ≥ 1 identity field and ≥ 1 `trust_level` field; `clauses` has ≥ 1 item — a boundary that admits every crossing is not a boundary; `operations` forbidden | | secret | `Secret` | `fields` | `clauses`, `operations`, `lifecycle: SecretLifecycle`, `material_ref: SemanticId`, `custodian: SemanticId`, `classification: ConfidentialityLevel` | `fields` has ≥ 1 item, ≥ 1 identity field, and 0 defaulted fields; `relations` forbidden | | encryption_key | `EncryptionKey` | `fields` | `clauses`, `operations`, `lifecycle: SecretLifecycle`, `material_ref: SemanticId`, `custodian: SemanticId`, `algorithm: Identifier`, `rotation: string` (ISO 8601 duration) | `fields` has ≥ 1 item, ≥ 1 identity field, and 0 defaulted fields; `relations` forbidden | | jwt_claim | `JwtClaim` | `fields` | `clauses`, `classification: ConfidentialityLevel`, `issuer: SemanticId` | `fields` has ≥ 1 item, ≥ 1 identity field, and 0 defaulted fields; `operations` and `relations` forbidden | @@ -91,12 +91,12 @@ semantic-core release that renders `identity: false` is a breaking change to the - The TypeSpec source SHALL express every positive item rule through the official emitter's `@contains` decorator over an open marker model: `@contains(IdentityField)` for "≥ 1 identity field", `@contains(OccurrenceField)` for "≥ 1 occurrence field". - The TypeSpec source SHALL express every negative item rule ("0 identity fields", "0 defaulted fields") as `items: { not: { $ref: } }` rather than as a counting predicate, for the reason [FR-006](./FR-006-security-safe-declarations.md) gives: a counting predicate inverts under a validator that does not implement `minContains`/`maxContains`, and `items`/`not` cannot. - Because JSON Schema admits one `contains` per array, the TypeSpec source SHALL carry every predicate beyond the first on that array as an `@extension("allOf", …)` clause referencing the marker file (`DefaultedField.json`, `OccurrenceField.json`, `IdentityField.json`, `SeverityField.json`, and the rest); the generator normalizes those relative `$ref`s per FR-002. -- Every cross-reference a declaration makes (`type.target`, `RelationDecl.target`, `owner`, `responsible`, `resource`, `grants`, `includes`, `exploits`, `affects`, `arises_from`, `traces_to`, `exposes`, `crosses`, `material_ref`, `custodian`, `issuer`, `binds`, `binds_flow`, `governs`, `mitigates`, `source`) SHALL be a `SemanticId` or `KernelScalar` per semantic-core, so a bare token is rejected by the schema; resolution against the bundle, and the placeholder `ix:////unresolved/` with its `semantic.unresolved-type` finding, exist today for `type.target` only (quire-rs FR-070). No published mapping covers the other keys: `agent-ix/quoin#335` is the ticket that would own one, and its body scopes itself to the ten `spec-objects-business` keys and names none of these, so this module records the gap rather than assuming that ticket will close it. +- Every cross-reference a declaration makes (`type.target`, `RelationDecl.target`, `owner`, `responsible`, `resource`, `grants`, `includes`, `exploits`, `affects`, `arises_from`, `traces_to`, `exposes`, `material_ref`, `custodian`, `issuer`, `binds`, `binds_flow`, `governs`, `mitigates`, `source`) SHALL be a `SemanticId` or `KernelScalar` per semantic-core, so a bare token is rejected by the schema; resolution against the bundle, and the placeholder `ix:////unresolved/` with its `semantic.unresolved-type` finding, exist today for `type.target` only (quire-rs FR-070). No published mapping covers the other keys: `agent-ix/quoin#335` is the ticket that would own one, and its body scopes itself to the ten `spec-objects-business` keys and names none of these, so this module records the gap rather than assuming that ticket will close it. - The `MfaMethod`, `Threat`, `EncryptionKey` and `CsrfToken` schemas SHALL each declare one optional string key for what its skeleton authors in frontmatter, because the engine projects every authored frontmatter key into the record that the closed schema validates: `kind` (the MFA method label, for example `totp`), `vector` (the attack-vector text), `rotation` and `rotation_window`. No other model gains a key from this rule. - The `MfaMethod` and `Threat` schemas SHALL require `kind` and `vector` to be non-empty strings when present. - The `EncryptionKey` and `CsrfToken` schemas SHALL require `rotation` and `rotation_window` to be ISO 8601 durations (for example `P90D`, `PT30M`) when present, refusing any other text. - Each schema SHALL describe the declared shape only, never a runtime occurrence (a session, an emitted audit record, a scanner hit), which is why `AuditEvent` carries no identity field: occurrence identity belongs to the runtime record, not the declaration. -- Where a key is declared but the current extractor does not populate it (`relations`, `classification`, `owner`, `grants`, `includes`, `resource`, `verb`, `factor_kind`, `stride_category`, `severity`, `cve_id`, `likelihood`, `impact`, `status`, `level`, `values`, `trust_level`, `crosses`, `exposes`, `lifecycle`, `material_ref`, `custodian`, `algorithm`, `issuer`, `binds`, `binds_flow`, `governs`, `mitigates`, `steps`, `source`, `affects`, `arises_from`, `traces_to`, `exploits`, `allowed_origins`) the key SHALL be optional, so a record produced by today's extractor validates and a future extractor can fill it without a schema change. +- Where a key is declared but the current extractor does not populate it (`relations`, `classification`, `owner`, `grants`, `includes`, `resource`, `verb`, `factor_kind`, `stride_category`, `severity`, `cve_id`, `likelihood`, `impact`, `status`, `level`, `values`, `crosses`, `exposes`, `lifecycle`, `material_ref`, `custodian`, `algorithm`, `issuer`, `binds`, `binds_flow`, `governs`, `mitigates`, `steps`, `source`, `affects`, `arises_from`, `traces_to`, `exploits`, `allowed_origins`) the key SHALL be optional, so a record produced by today's extractor validates and a future extractor can fill it without a schema change. - Eight of the optional graded keys SHALL additionally be expressed as an item rule over `fields`, which the extractor does populate — `stride_category`, `severity`, `likelihood`, `impact`, `status`, `level`, `trust_level` and `effectiveness`, one open marker model each — so a threat states its STRIDE class as a row and a control its effectiveness as a row, and the obligation is enforced today rather than deferred. - The remaining graded keys (`lifecycle`, `classification`, `factor_kind`) SHALL carry no item rule: they record how something is held rather than how bad it is, and no refusal in this module turns on them. They are declared, typed, and optional; nothing else is claimed for them. - The test suite SHALL verify every criterion over a key the extractor does not populate against a hand-built JSON record rather than an extracted one, naming that limitation in the test itself, so that no row claims extraction evidence it does not have. diff --git a/spec/functional/FR-024-trust-boundary-trust-level.md b/spec/functional/FR-024-trust-boundary-trust-level.md new file mode 100644 index 0000000..c09d08a --- /dev/null +++ b/spec/functional/FR-024-trust-boundary-trust-level.md @@ -0,0 +1,50 @@ +--- +id: FR-024 +title: "A trust_boundary states its trust level as a required frontmatter value" +type: FR +relationships: + - target: "ix://agent-ix/spec-objects-security/FR-004" + type: depends_on + - target: "ix://agent-ix/spec-objects-security/FR-023" + type: depends_on + - target: "ix://agent-ix/spec-objects-architecture/FR-014" + type: depends_on +--- +# FR-024: A trust_boundary states its trust level as a required frontmatter value + +## Description + +A trust boundary identifies the zone on its far side and the trust given to that zone. A field row in `## Properties` describes a value but does not supply the value to consumers. The authored boundary therefore states `trust_level` in frontmatter, where the extractor places it in the typed `TrustBoundary` record. Connection crossings are relationship edges owned by the connection and are specified by spec-objects-architecture FR-014; `TrustBoundary` carries no `crosses` record key. + +## Behavior + +- Each `trust_boundary` SHALL author frontmatter `trust_level` with exactly one value from the existing `TrustLevel` enumeration: `untrusted`, `semi_trusted`, `trusted`, or `unknown`. +- `trust_level` SHALL be required in the frontmatter locator and SHALL be extracted to the `trust_level` record key. +- A missing `trust_level` SHALL make `quire validate` fail with a diagnostic naming `trust_level`. +- An out-of-vocabulary value SHALL make `quire validate` fail with a diagnostic naming the value. +- The emitted `TrustBoundary` schema SHALL admit the typed `trust_level` key and SHALL reject the retired `crosses` record key. Crossing relationships are authored on connections. +- The existing `fields` row for `trust_level` remains required by FR-004-AC-10 and its `TrustLevelField` item rule. The row and the frontmatter value serve distinct purposes: the row documents a boundary property, while the frontmatter value is the extracted machine-readable value. +- The trust-boundary skeleton SHALL include a valid frontmatter `trust_level`. +- No trust level is inferred from a boundary name, diagram, relation, or property row. + +## Acceptance Criteria + +| ID | Criteria | Verification | +|----|----------|--------------| +| FR-024-AC-1 | A `trust_boundary` with a valid `trust_level` frontmatter value passes `quire validate`. | Test | +| FR-024-AC-2 | A `trust_boundary` without frontmatter `trust_level` fails validation and names the missing field. | Test | +| FR-024-AC-3 | A `trust_level` outside `untrusted`, `semi_trusted`, `trusted`, and `unknown` fails validation and names the value. | Test | +| FR-024-AC-4 | `trust_level: untrusted` is extracted as the record key `trust_level` with value `untrusted`. | Test | +| FR-024-AC-5 | The `TrustBoundary` schema rejects a record containing `crosses`, while a valid typed `trust_level` is admitted; the `TrustLevelField` item rule remains. | Test | +| FR-024-AC-6 | The trust-boundary skeleton carries valid frontmatter `trust_level` and validates. | Test | +| FR-024-AC-7 | A trust level is not derived from the `## Properties` row: removing frontmatter leaves no extracted `trust_level` and validation fails. | Test | + +## Dependencies + +- FR-004 defines the `TrustBoundary` record shape and the `TrustLevelField` item rule. +- FR-023 defines the security module's typed frontmatter extraction pattern for authored enum values. +- spec-objects-architecture FR-014 owns the connection-to-boundary `crosses` edge and its relation report. + +## Licensing + +Use generic terms only. Do not name or quote paywalled standards. diff --git a/spec/functional/index.md b/spec/functional/index.md index 04778e3..1df31f7 100644 --- a/spec/functional/index.md +++ b/spec/functional/index.md @@ -28,6 +28,7 @@ description: "Index of artifacts in this directory." * [FR-021: Default source profiles list only publishable clause sets and degrade while a pack is missing](./FR-021-default-source-profiles.md) * [FR-022: A SOUP record carries optional person-authored VEX statements](./FR-022-vex-statements-on-soup-records.md) * [FR-023: A data_classification states its regulated-data categories as a list separate from its confidentiality level](./FR-023-regulated-data-categories-on-data-classification.md) +* [FR-024: A trust_boundary states its trust level as a required frontmatter value](./FR-024-trust-boundary-trust-level.md) * [FR-025: A control states safeguards and protects a connection or data_sink, and the module ships the remediation vocabulary and starter flow rules](./FR-025-control-safeguards-and-starter-flow-rules.md) * [FR-026: A threat_model states its method, who led it, what it covers and which threats it considers](./FR-026-threat-model-object.md) * [FR-027: A threat targets model elements, a risk records a named person's decision, and the module ships the threat-model completeness flow rules](./FR-027-threat-treatment-and-completeness-rules.md) diff --git a/spec/tests.md b/spec/tests.md index 43fe55a..1ea66ee 100644 --- a/spec/tests.md +++ b/spec/tests.md @@ -132,7 +132,7 @@ the Python package only: neither Node script is measured by it. | FR-001 | FR-001-AC-1..4 | TC-001..TC-004, TC-006..TC-014, TC-015..TC-017 | ✅ | | FR-002 | FR-002-AC-1..4, FR-002-AC-6..11, FR-002-CON-1..4 | TC-020..TC-035 | ✅ | | FR-003 | FR-003-AC-1, FR-003-AC-4..7, FR-003-CON-1 | TC-040, TC-041, TC-044..TC-046, TC-115 | 🚧 AC-5 runs install, listing, and restoration in a temporary HOME; AC-7 independently checks both engine gates; AC-6's naming half is an expected failure | -| FR-004 | FR-004-AC-1..15, FR-004-CON-1..2 | TC-050..TC-063, TC-218, TC-219 | 🚧 | +| FR-004 | FR-004-AC-1..15, FR-004-CON-1..2 | TC-050..TC-063, TC-218..TC-220 | 🚧 | | FR-005 | FR-005-AC-1..9, FR-005-CON-1..3 | TC-080..TC-090 | 🚧 | | FR-006 | FR-006-AC-1..7, FR-006-CON-1..2 | TC-070..TC-077 | ✅ | | FR-007 | FR-007-AC-1..2 | TC-091, TC-092 | ✅ | @@ -151,6 +151,7 @@ the Python package only: neither Node script is measured by it. | FR-016 | FR-016-AC-1..9 | TC-197..TC-203, TC-205 | ✅ AC-8 is also exercised with NFR-003-AC-2 by TC-205; AC-9 covers the decision vocabulary and conditional RFC 3339 expiry in TC-199 | | NFR-003 | NFR-003-AC-1..2, NFR-003-M-1..2 | TC-204, TC-205 | ✅ | | FR-023 | FR-023-AC-1..10 | TC-206..TC-217 | ✅ AC-7 is an inspection | +| FR-024 | FR-024-AC-1..7 | TC-220..TC-223 | ✅ | | NFR-004 | NFR-004-AC-1..3, NFR-004-M-1..2 | TC-207, TC-211, TC-213, TC-214 | ✅ | | FR-017 | FR-017-AC-1..7 | TC-175..TC-186 | ✅ `quire clauses` loads strictly and cannot load the real module (the object-type edges come from another module), so TC-181, TC-183, TC-185 and TC-186 evaluate a module that holds only the set files | @@ -339,6 +340,10 @@ the Python package only: neither Node script is measured by it. | TC-217 | A section beside frontmatter is a `frontmatter-owned` warning and the record keeps only the frontmatter list | Unit | P0 | FR-023-AC-10 | ✅ | | TC-218 | The authored `kind`, `vector`, `rotation` and `rotation_window` keys are optional; non-empty text and ISO 8601 durations are accepted, and empty text, bare, repeated or out-of-order durations and non-strings are refused | Unit | P0 | FR-004-AC-15 | ✅ | | TC-219 | A malformed `rotation` or `rotation_window` in an authored document fails with `semantic.record-invalid` | Unit | P0 | FR-004-AC-15 | ✅ | +| TC-220 | TrustBoundary admits a typed trust_level and refuses the retired crosses record key | Unit | P0 | FR-004-AC-10, FR-024-AC-5 | ✅ | +| TC-221 | trust_boundary requires frontmatter trust_level and refuses an absent or out-of-vocabulary value | Unit | P0 | FR-024-AC-1, FR-024-AC-2, FR-024-AC-3 | ✅ | +| TC-222 | trust_level is extracted from frontmatter to the typed record key and is not derived from the Properties row | Unit | P0 | FR-024-AC-4, FR-024-AC-7 | ✅ | +| TC-223 | the trust_boundary skeleton states an authored trust_level and validates | Unit | P0 | FR-024-AC-6 | ✅ | ## Test Environment @@ -349,7 +354,7 @@ row here can be reported green without the engine under test. This is the disposition of `agent-ix/spec-objects-security#10`. Rows over the record keys the extractor does not populate (`severity`, -`likelihood`, `impact`, `status`, `level`, `trust_level`, `stride_category`, +`likelihood`, `impact`, `status`, `level`, `stride_category`, `effectiveness`, `lifecycle`, `material_ref`, `relations`, and the cross-reference lists) are verified against hand-built records, not extracted ones — TC-053, TC-054, TC-059, TC-063, TC-073 in particular — and their tests diff --git a/spec_objects_security/manifest.yaml b/spec_objects_security/manifest.yaml index 41debf1..68348ee 100644 --- a/spec_objects_security/manifest.yaml +++ b/spec_objects_security/manifest.yaml @@ -269,6 +269,10 @@ object_types: id: *id001 title: *id002 type: *id003 + trust_level: + from: frontmatter_field + path: [trust_level] + required: true boundary: from: code_block after_heading: Boundary diff --git a/spec_objects_security/schemas/TrustBoundary.json b/spec_objects_security/schemas/TrustBoundary.json index a61b4bb..69c9e65 100644 --- a/spec_objects_security/schemas/TrustBoundary.json +++ b/spec_objects_security/schemas/TrustBoundary.json @@ -28,12 +28,6 @@ }, "trust_level": { "$ref": "https://schemas.agent-ix.org/agent-ix/spec-objects-security/TrustLevel.json" - }, - "crosses": { - "type": "array", - "items": { - "$ref": "https://schemas.agent-ix.org/semantic-core/0.4.4/SemanticId.json" - } } }, "required": [ diff --git a/spec_objects_security/skeletons/trust_boundary.md b/spec_objects_security/skeletons/trust_boundary.md index 5db9937..a439558 100644 --- a/spec_objects_security/skeletons/trust_boundary.md +++ b/spec_objects_security/skeletons/trust_boundary.md @@ -3,11 +3,12 @@ id: BOUND-001 title: "TenantNetworkEdge" type: trust_boundary object: trust_boundary +trust_level: untrusted --- + +| FND | Severity | Check | Summary | +| --- | --- | --- | --- | +| FND-001 | high | soundness | The new _extract helper duplicates tests/test_regulated_categories.py::_extract: it loads schema documents, composes module mappings and source identity, then invokes extract_semantic with dataSchema and bodyExtraction. This creates two implementations that can drift; reuse or generalize the existing helper. | + ++++ [reviewer data] + +```yaml +clean: false +scope: +- id: tests/test_trust_boundary.py::_extract + path: tests/test_trust_boundary.py + role: examined + excerpt: Semantic extraction helper loads schemas and invokes extract_semantic with + the module mappings, source identity, data schema, and body extraction. +- id: tests/test_regulated_categories.py::_extract + path: tests/test_regulated_categories.py + role: context_only + excerpt: Existing helper performs the same semantic extraction setup with schema + documents, module mappings, source identity, and bodyExtraction. +findings: +- fnd: FND-001 + severity: high + confidence: high + method: code-review + check: soundness + artifact_id: tests/test_trust_boundary.py::_extract + related_ids: + - tests/test_regulated_categories.py::_extract + path: tests/test_trust_boundary.py + lines: 20-42 + excerpt: The trust_level frontmatter extraction path has separate coverage in test_trust_boundary.py. + finding: 'The new _extract helper duplicates tests/test_regulated_categories.py::_extract: + it loads schema documents, composes module mappings and source identity, then + invokes extract_semantic with dataSchema and bodyExtraction. This creates two + implementations that can drift; reuse or generalize the existing helper.' +bindings: [] +``` + ++++ \ No newline at end of file diff --git a/reviews/SR-1614-gap-analysis.comment.md b/reviews/SR-1614-gap-analysis.comment.md new file mode 100644 index 0000000..f551737 --- /dev/null +++ b/reviews/SR-1614-gap-analysis.comment.md @@ -0,0 +1,46 @@ + + + + ++++ [reviewer data] + +```yaml +clean: true +scope: +- id: FR-024-AC-1 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: A trust_boundary with a valid trust_level frontmatter value passes quire + validate. +- id: FR-024-AC-2 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: A trust_boundary without frontmatter trust_level fails validation and names + the missing field. +- id: FR-024-AC-3 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: A trust_level outside the allowed enum fails validation and names the value. +- id: FR-024-AC-4 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: 'trust_level: untrusted is extracted as the record key trust_level with + value untrusted.' +- id: FR-024-AC-5 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: The TrustBoundary schema rejects crosses, admits a valid typed trust_level, + and keeps the TrustLevelField item rule. +- id: FR-024-AC-6 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: The trust-boundary skeleton carries valid frontmatter trust_level and validates. +- id: FR-024-AC-7 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: Removing frontmatter leaves no extracted trust_level and validation fails. +findings: [] +bindings: [] +``` + ++++ \ No newline at end of file diff --git a/reviews/SR-1615-spec-review-base.comment.md b/reviews/SR-1615-spec-review-base.comment.md new file mode 100644 index 0000000..32b856c --- /dev/null +++ b/reviews/SR-1615-spec-review-base.comment.md @@ -0,0 +1,20 @@ + + + + ++++ [reviewer data] + +```yaml +clean: true +scope: +- id: FR-024 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: 'Each trust_boundary SHALL author frontmatter trust_level with exactly + one value from the existing TrustLevel enumeration: untrusted, semi_trusted, trusted, + or unknown.' +findings: [] +bindings: [] +``` + ++++ \ No newline at end of file diff --git a/reviews/SR-1616-spec-review-ears-conformance.comment.md b/reviews/SR-1616-spec-review-ears-conformance.comment.md new file mode 100644 index 0000000..791a738 --- /dev/null +++ b/reviews/SR-1616-spec-review-ears-conformance.comment.md @@ -0,0 +1,20 @@ + + + + ++++ [reviewer data] + +```yaml +clean: true +scope: +- id: FR-024 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: 'Each trust_boundary SHALL author frontmatter trust_level with exactly + one value from the existing TrustLevel enumeration: untrusted, semi_trusted, trusted, + or unknown.' +findings: [] +bindings: [] +``` + ++++ \ No newline at end of file diff --git a/reviews/SR-1617-spec-review-integrity.comment.md b/reviews/SR-1617-spec-review-integrity.comment.md new file mode 100644 index 0000000..6fc9c00 --- /dev/null +++ b/reviews/SR-1617-spec-review-integrity.comment.md @@ -0,0 +1,19 @@ + + + + ++++ [reviewer data] + +```yaml +clean: true +scope: +- id: FR-024-AC-5 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: The TrustBoundary schema rejects a record containing crosses, while a valid + typed trust_level is admitted; the TrustLevelField item rule remains. +findings: [] +bindings: [] +``` + ++++ \ No newline at end of file diff --git a/reviews/SR-1619-spec-review-dependency.comment.md b/reviews/SR-1619-spec-review-dependency.comment.md new file mode 100644 index 0000000..adf53ae --- /dev/null +++ b/reviews/SR-1619-spec-review-dependency.comment.md @@ -0,0 +1,43 @@ + + +| FND | Severity | Check | Summary | +| --- | --- | --- | --- | +| FND-001 | medium | soundness | FR-024 depends on architecture FR-014 even though trust-level extraction, validation, and the security record shape stand independently. FR-014 depends on FR-024 to report trust_level, so this reverse depends_on edge creates a circular dependency. Remove FR-024 → FR-014; state that connection edges are architecture-owned without making trust-level extraction depend on the crossing feature. | + ++++ [reviewer data] + +```yaml +clean: false +scope: +- id: FR-024 + path: spec/functional/FR-024-trust-boundary-trust-level.md + role: examined + excerpt: 'relationships: depends_on FR-004; depends_on FR-023; depends_on architecture + FR-014.' +- id: FR-014 + path: ix://agent-ix/spec-objects-architecture/FR-014 + role: context_only + excerpt: FR-014 reports each crossing target with its trust_level, requiring the + TrustBoundary trust_level field. +findings: +- fnd: FND-001 + severity: medium + confidence: high + method: spec-review/dependency + check: soundness + artifact_id: FR-024 + related_ids: + - FR-014 + path: spec/functional/FR-024-trust-boundary-trust-level.md + lines: 2-9 + excerpt: "relationships:\n - target: \"ix://agent-ix/spec-objects-architecture/FR-014\"\ + \n type: depends_on" + finding: FR-024 depends on architecture FR-014 even though trust-level extraction, + validation, and the security record shape stand independently. FR-014 depends + on FR-024 to report trust_level, so this reverse depends_on edge creates a circular + dependency. Remove FR-024 → FR-014; state that connection edges are architecture-owned + without making trust-level extraction depend on the crossing feature. +bindings: [] +``` + ++++ \ No newline at end of file From be5b71355ad9f731ad5a2862a1d6e66e6c7561d8 Mon Sep 17 00:00:00 2001 From: Agent IX Date: Sun, 15 Nov 2026 18:15:09 -0800 Subject: [PATCH 7/7] Record PLAT-1532 security review dispositions --- reviews/plat-1532-code-review.md | 12 ++++++++++++ reviews/plat-1532-dependency.md | 12 ++++++++++++ 2 files changed, 24 insertions(+) diff --git a/reviews/plat-1532-code-review.md b/reviews/plat-1532-code-review.md index 1b17499..0f49887 100644 --- a/reviews/plat-1532-code-review.md +++ b/reviews/plat-1532-code-review.md @@ -36,3 +36,15 @@ Observed finding: The new _extract helper repeats the semantic extraction setup - `tests/test_trust_boundary.py::_extract` (examined): “Semantic extraction helper loads schemas and invokes extract_semantic with the module mappings, source identity, data schema, and body extraction.” - `tests/test_regulated_categories.py::_extract` (context_only): “Existing helper performs the same semantic extraction setup with schema documents, module mappings, source identity, and bodyExtraction.” + +## Dispositions + +| FND | outcome | reason | +| --- | --- | --- | +| FND-001 | fixed | The trust-boundary test now reuses the generalized `_extract` helper from `test_regulated_categories.py`. | + +## Dispositions + +| FND | outcome | reason | +| --- | --- | --- | +| FND-001 | fixed | The extraction logic is generalized in `test_regulated_categories.py::_extract`, and `test_trust_boundary.py` now reuses it with the trust_boundary schema and bundle. | diff --git a/reviews/plat-1532-dependency.md b/reviews/plat-1532-dependency.md index 0cb9e54..bc56286 100644 --- a/reviews/plat-1532-dependency.md +++ b/reviews/plat-1532-dependency.md @@ -36,3 +36,15 @@ Observed finding: FR-024 depends on architecture FR-014 even though trust-level - `FR-024` (examined): “relationships: depends_on FR-004; depends_on FR-023; depends_on architecture FR-014.” - `FR-014` (context_only): “FR-014 reports each crossing target with its trust_level, requiring the TrustBoundary trust_level field.” + +## Dispositions + +| FND | outcome | reason | +| --- | --- | --- | +| FND-001 | fixed | FR-024 no longer has a `depends_on` edge to architecture FR-014; the text says extraction is independent of the crossing edge. | + +## Dispositions + +| FND | outcome | reason | +| --- | --- | --- | +| FND-001 | fixed | FR-024 no longer declares a `depends_on` edge to FR-014 and explicitly says trust-level extraction does not depend on the crossing edge. |