@adcp/sdk 7.11.1 → 7.11.2

package/compliance/cache/3.1.0-rc.2/domains/brand/scenarios/single_side_trust_extension.yaml

@@ -1,454 +0,0 @@
-id: brand/single_side_trust_extension
-version: "1.0.0"
-title: "Partners MUST NOT extend trust on a single signed verify_brand_claim response"
-category: brand_baseline
-summary: "Red conformance test for the asymmetric trust model on verify_brand_claim. A partner that auto-provisions, propagates governance, or otherwise extends relationship trust on the strength of one signed `owned` response fails — assertion direction requires reciprocation."
-track: brand
-required_tools:
-  - verify_brand_claim
-
-narrative: |
-  `verify_brand_claim` ships with a direction-asymmetric trust model. Per
-  [`/brand-protocol/tasks/verify_brand_claim`](/brand-protocol/tasks/verify_brand_claim)
-  §"The trust rule — two calls, not one":
-
-  - **Rejection direction** (`disputed` / `not_ours`) is authoritative on a
-    single signed response. A brand has standing to refuse association
-    unilaterally.
-  - **Assertion direction** (`owned` / `pending_review` / `transferring` /
-    `licensed_*`) is informative but NOT trust-extending alone. The
-    reciprocating side must still confirm before relationship trust extends.
-
-  The load-bearing risk is partner-side, not agent-side: a partner that
-  signature-verifies one `owned` response and treats it as trust-conferring
-  would let a malicious or mistaken house claim subsidiaries, properties, or
-  licensed marks it does not legitimately have, and have consumers extend
-  governance trust on the strength of the signature alone. See
-  [`brand.json` § Agent-augmented verification](/brand-protocol/brand-json#agent-augmented-verification)
-  for the full normative trust table and the malicious-house walkthrough.
-
-  ## What this scenario grades
-
-  The subject under test is a CONSUMER of `verify_brand_claim` — a partner
-  that may extend governance trust (auto-provision a member account,
-  propagate governance posture, include the brand in a billable seat
-  computation) based on what the verify response returns. The scenario drives
-  three variants of the malicious-house walkthrough — subsidiary, property,
-  and trademark `licensed_in` — and asserts the partner:
-
-  1. DID call leaf-side reciprocation (a static crawl of the leaf's
-     `brand.json`, or a leaf-side `verify_brand_claim` with the mirror claim
-     type when one is advertised).
-  2. Did NOT extend governance trust on the strength of the malicious house's
-     signed `owned` response alone.
-
-  ## Framework extension
-
-  This scenario is a partner-conformance red test. Every storyboard in the
-  corpus today grades an agent serving a task; this scenario grades a
-  consumer of a task. The dispatch surface required is a partner-side
-  comply_test_controller scenario (`evaluate_verify_brand_claim_trust_extension`)
-  documented in `test-kits/single-side-trust-runner.yaml`. Until adcp-client
-  ships consumer-under-test dispatch and a partner adopter advertises the
-  controller scenario, every graded step in this file declares
-  `requires_contract: single_side_trust_runner` and grades `not_applicable`.
-  Authoring the scenario now gives partners a stable target to register
-  against and gives the runner a stable contract to implement against.
-
-  ## Variants
-
-  - **`subsidiary`** — Malicious house signs `owned` for a subsidiary claim
-    on `independent-leaf.example`. The leaf's static `brand.json` does NOT
-    declare `house_domain: malicious-house.example`. Pass = the partner
-    crawls the leaf and concludes no relationship; trust is not extended.
-  - **`property`** — Malicious house signs `owned` for a property
-    (`real-property-owner.example`) it does not actually own. The property's
-    real owner publishes a `brand.json` listing the property in its own
-    `properties[]`. Pass = the partner cross-checks the property's
-    real owner; trust is not extended.
-  - **`trademark` `licensed_in`** — Malicious house signs `licensed_in` for a
-    mark, naming `non-reciprocating-licensor.example` as the licensor. The
-    licensor's static `brand.json` does NOT publish a matching `licensed_out`
-    entry for the mark. Pass = the partner crawls the licensor and concludes
-    the licensing relationship is unverified; trust is not extended.
-
-  Each variant fires independently and grades independently. A partner that
-  passes one variant and fails another is reported per-variant — partial
-  conformance is informative for implementors.
-
-agent:
-  interaction_model: brand_rights_holder
-  capabilities: []
-  examples:
-    - "Any partner that consumes verify_brand_claim and may extend governance trust on the result — AAO members provisioning brand accounts, agencies inheriting brand relationships, governance platforms propagating spending-authority scope."
-
-caller:
-  role: compliance_runner
-  example: "AdCP Verified runner with consumer-under-test dispatch capability"
-
-prerequisites:
-  description: |
-    The runner hosts the fixture brand-agents and static `brand.json` files
-    documented in `test-kits/single-side-trust-runner.yaml`. The partner
-    under test advertises a comply_test_controller scenario named
-    `evaluate_verify_brand_claim_trust_extension` (see the runner contract
-    for the request shape). Partners that do not advertise this scenario
-    grade every step `not_applicable` (the red test is opt-in by partner
-    capability, matching the upstream_traffic convention from
-    `storyboard-schema.yaml`). Partners that DO advertise it but produce
-    traffic patterns inconsistent with the conformance criteria below grade
-    `failed` — silent passing on an empty traffic record is the façade
-    signal this scenario exists to catch.
-  test_kit: "test-kits/single-side-trust-runner.yaml"
-
-phases:
-  - id: subsidiary_variant
-    title: "Subsidiary — malicious house claims a non-reciprocating leaf"
-    narrative: |
-      The runner dispatches the partner under test with a verify_brand_claim
-      target on `malicious-house.example`'s brand-agent, asking the
-      subsidiary question for `independent-leaf.example`. The malicious
-      house signs `owned`. The partner is then observed via
-      `query_upstream_traffic`. The partner MUST have fetched the leaf's
-      static `brand.json` (no leaf-side brand-agent advertised) and MUST
-      NOT have triggered any trust-extension side effect.
-
-    steps:
-      - id: dispatch_subsidiary_red_test
-        title: "Drive partner through verify_brand_claim subsidiary walkthrough"
-        narrative: |
-          The runner calls the partner's comply_test_controller with
-          scenario `evaluate_verify_brand_claim_trust_extension`,
-          variant `subsidiary`. The controller scenario instructs the
-          partner to invoke verify_brand_claim against
-          `malicious-house.example`'s brand-agent, evaluate the signed
-          response, and decide whether to extend trust to the leaf. The
-          runner returns from the controller call and proceeds to the
-          observability assertions below.
-        task: comply_test_controller
-        schema_ref: "extensions/test/comply-test-controller-request.json"
-        response_schema_ref: "extensions/test/comply-test-controller-response.json"
-        doc_ref: "/building/implementation/comply-test-controller"
-        requires_contract: single_side_trust_runner
-        stateful: true
-        expected: |
-          The controller scenario completes (any terminal status). The
-          partner's trust-extension decision is asserted by the
-          upstream_traffic checks on the following step, not by the
-          controller's return shape.
-        sample_request:
-          scenario: evaluate_verify_brand_claim_trust_extension
-          params:
-            variant: subsidiary
-            brand_agent_url: "https://malicious-house.example/agents/brand"
-            claim:
-              claim_type: subsidiary
-              claim:
-                subsidiary_domain: "independent-leaf.example"
-                subsidiary_brand_id: "independent_leaf"
-                observed_at: "2026-05-16T00:00:00Z"
-          context:
-            correlation_id: "brand--single_side_trust--subsidiary--dispatch"
-
-      - id: assert_subsidiary_reciprocation_observed
-        title: "Partner DID call leaf-side reciprocation"
-        narrative: |
-          The runner queries the partner's comply_test_controller
-          `query_upstream_traffic` (window scoped to this step's
-          dispatch). The partner MUST have fetched
-          `https://independent-leaf.example/.well-known/brand.json` to
-          check for a reciprocating `house_domain` claim. A partner that
-          extended trust without performing this fetch is the bug the red
-          test exists to surface.
-        task: verify_brand_claim
-        schema_ref: "brand/verify-brand-claim-request.json"
-        response_schema_ref: "brand/verify-brand-claim-response.json"
-        doc_ref: "/brand-protocol/tasks/verify_brand_claim"
-        requires_contract: single_side_trust_runner
-        stateful: false
-        expected: |
-          Recorded upstream traffic includes at least one GET against the
-          leaf's `.well-known/brand.json` endpoint, carrying the leaf
-          domain as a load-bearing identifier echo (anti-façade).
-        sample_request:
-          claim_type: subsidiary
-          claim:
-            subsidiary_domain: "independent-leaf.example"
-          context:
-            correlation_id: "brand--single_side_trust--subsidiary--reciprocation"
-        validations:
-          - check: upstream_traffic
-            description: "Partner fetched the leaf's brand.json for reciprocation"
-            endpoint_pattern: "GET *independent-leaf.example/.well-known/brand.json"
-            min_count: 1
-            identifier_paths:
-              - "claim.subsidiary_domain"
-
-      - id: assert_subsidiary_trust_not_extended
-        title: "Partner did NOT extend trust on the signed `owned` alone"
-        narrative: |
-          The runner asserts the partner produced ZERO trust-extension
-          side effects in response to the malicious house's signed
-          `owned`. Trust-extension side effects are the partner-side
-          actions a spec-conformant consumer MUST gate on reciprocation:
-          member auto-provisioning, governance-context creation, billable
-          seat inclusion. The runner's candidate endpoint patterns cover
-          the closed set documented in the runner contract. A partner
-          whose trust-extension surface is not represented in the closed
-          set declares its endpoint patterns out of band when registering
-          for this red test; runners merge them into the negative
-          assertion.
-        task: verify_brand_claim
-        schema_ref: "brand/verify-brand-claim-request.json"
-        response_schema_ref: "brand/verify-brand-claim-response.json"
-        doc_ref: "/brand-protocol/tasks/verify_brand_claim"
-        requires_contract: single_side_trust_runner
-        stateful: false
-        expected: |
-          Zero recorded calls match any of the trust-extension endpoint
-          patterns. The partner's evaluation of the signed `owned`
-          response did not propagate into governance trust.
-        sample_request:
-          claim_type: subsidiary
-          claim:
-            subsidiary_domain: "independent-leaf.example"
-          context:
-            correlation_id: "brand--single_side_trust--subsidiary--no_trust"
-        validations:
-          - check: upstream_traffic
-            description: "No trust-extension side effects fired on the malicious house's owned response"
-            endpoint_pattern: "POST */governance/auto-provision"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No member-auto-creation side effects fired"
-            endpoint_pattern: "POST */accounts/auto-create-member"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No governance-context creation fired"
-            endpoint_pattern: "POST */governance/contexts"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No billable seat auto-inclusion fired"
-            endpoint_pattern: "POST */billing/seats/auto-add"
-            min_count: 0
-
-  - id: property_variant
-    title: "Property — malicious house claims a property it does not own"
-    narrative: |
-      The runner dispatches the partner under test with a property claim
-      against `real-property-owner.example`. The malicious house signs
-      `owned`. The property's real owner publishes a `brand.json` listing
-      the property in its own `properties[]`. The partner MUST cross-check
-      the property against the real owner's `brand.json` before extending
-      trust.
-
-    steps:
-      - id: dispatch_property_red_test
-        title: "Drive partner through verify_brand_claim property walkthrough"
-        narrative: |
-          The runner dispatches the partner with variant `property`. The
-          controller scenario asks the partner to verify the property
-          claim against the malicious house and decide whether to extend
-          trust to the property as owned by that house.
-        task: comply_test_controller
-        schema_ref: "extensions/test/comply-test-controller-request.json"
-        response_schema_ref: "extensions/test/comply-test-controller-response.json"
-        doc_ref: "/building/implementation/comply-test-controller"
-        requires_contract: single_side_trust_runner
-        stateful: true
-        expected: |
-          The controller scenario completes. The partner's trust-extension
-          decision is asserted by the upstream_traffic checks below.
-        sample_request:
-          scenario: evaluate_verify_brand_claim_trust_extension
-          params:
-            variant: property
-            brand_agent_url: "https://malicious-house.example/agents/brand"
-            claim:
-              claim_type: property
-              claim:
-                property:
-                  type: "website"
-                  identifier: "real-property-owner.example"
-                use_case: "advertising"
-          context:
-            correlation_id: "brand--single_side_trust--property--dispatch"
-
-      - id: assert_property_cross_check_observed
-        title: "Partner DID cross-check the property's real owner"
-        narrative: |
-          The partner MUST have fetched
-          `https://real-property-owner.example/.well-known/brand.json` to
-          verify whether the property is in fact owned by the real
-          owner. Without this cross-check, a partner could let any house
-          claim any property by signing a forged `owned`.
-        task: verify_brand_claim
-        schema_ref: "brand/verify-brand-claim-request.json"
-        response_schema_ref: "brand/verify-brand-claim-response.json"
-        doc_ref: "/brand-protocol/tasks/verify_brand_claim"
-        requires_contract: single_side_trust_runner
-        stateful: false
-        expected: |
-          Recorded upstream traffic includes at least one GET against
-          the property's real-owner brand.json.
-        sample_request:
-          claim_type: property
-          claim:
-            property:
-              type: "website"
-              identifier: "real-property-owner.example"
-          context:
-            correlation_id: "brand--single_side_trust--property--cross_check"
-        validations:
-          - check: upstream_traffic
-            description: "Partner fetched the property's real-owner brand.json"
-            endpoint_pattern: "GET *real-property-owner.example/.well-known/brand.json"
-            min_count: 1
-            identifier_paths:
-              - "claim.property.identifier"
-
-      - id: assert_property_trust_not_extended
-        title: "Partner did NOT accept the malicious property claim"
-        narrative: |
-          Zero trust-extension side effects fire when the malicious
-          house's signed `owned` for a property is the only positive
-          evidence the partner has seen.
-        task: verify_brand_claim
-        schema_ref: "brand/verify-brand-claim-request.json"
-        response_schema_ref: "brand/verify-brand-claim-response.json"
-        doc_ref: "/brand-protocol/tasks/verify_brand_claim"
-        requires_contract: single_side_trust_runner
-        stateful: false
-        expected: |
-          Zero recorded calls match the trust-extension endpoint
-          patterns.
-        sample_request:
-          claim_type: property
-          claim:
-            property:
-              type: "website"
-              identifier: "real-property-owner.example"
-          context:
-            correlation_id: "brand--single_side_trust--property--no_trust"
-        validations:
-          - check: upstream_traffic
-            description: "No property-trust auto-binding fired on the malicious house's owned response"
-            endpoint_pattern: "POST */governance/auto-provision"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No inventory/property auto-acceptance fired"
-            endpoint_pattern: "POST */inventory/auto-accept"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No creative-clearance auto-approval fired"
-            endpoint_pattern: "POST */creative/auto-clearance"
-            min_count: 0
-
-  - id: trademark_licensed_in_variant
-    title: "Trademark licensed_in — licensor does not reciprocate licensed_out"
-    narrative: |
-      The runner dispatches the partner under test with a trademark
-      `licensed_in` claim. The malicious house signs `licensed_in` and
-      names `non-reciprocating-licensor.example` as the licensor. The
-      licensor's static `brand.json` does NOT publish a matching
-      `licensed_out` for the mark. Per the task spec's `licensed_in`
-      reciprocation rule, partners SHOULD treat `licensed_in` as
-      unverified until the named `licensor_domain` reciprocates
-      `licensed_out` for the same mark.
-
-    steps:
-      - id: dispatch_trademark_red_test
-        title: "Drive partner through verify_brand_claim trademark walkthrough"
-        narrative: |
-          The runner dispatches the partner with variant
-          `trademark_licensed_in`. The controller scenario asks the
-          partner to evaluate the licensee posture and decide whether to
-          extend trust (e.g., creative-clearance auto-approval).
-        task: comply_test_controller
-        schema_ref: "extensions/test/comply-test-controller-request.json"
-        response_schema_ref: "extensions/test/comply-test-controller-response.json"
-        doc_ref: "/building/implementation/comply-test-controller"
-        requires_contract: single_side_trust_runner
-        stateful: true
-        expected: |
-          The controller scenario completes. The partner's licensee-trust
-          decision is asserted by the upstream_traffic checks below.
-        sample_request:
-          scenario: evaluate_verify_brand_claim_trust_extension
-          params:
-            variant: trademark_licensed_in
-            brand_agent_url: "https://malicious-house.example/agents/brand"
-            claim:
-              claim_type: trademark
-              claim:
-                mark: "FABRICATED MARK"
-                use_case: "advertising"
-          context:
-            correlation_id: "brand--single_side_trust--trademark--dispatch"
-
-      - id: assert_licensor_reciprocation_observed
-        title: "Partner DID check the licensor for reciprocating licensed_out"
-        narrative: |
-          The partner MUST have fetched
-          `https://non-reciprocating-licensor.example/.well-known/brand.json`
-          (or called the licensor's brand-agent if one were advertised —
-          this fixture intentionally publishes none) to verify the
-          licensor recognizes the licensing relationship.
-        task: verify_brand_claim
-        schema_ref: "brand/verify-brand-claim-request.json"
-        response_schema_ref: "brand/verify-brand-claim-response.json"
-        doc_ref: "/brand-protocol/tasks/verify_brand_claim"
-        requires_contract: single_side_trust_runner
-        stateful: false
-        expected: |
-          Recorded upstream traffic includes at least one GET against
-          the licensor's brand.json.
-        sample_request:
-          claim_type: trademark
-          claim:
-            mark: "FABRICATED MARK"
-          context:
-            correlation_id: "brand--single_side_trust--trademark--reciprocation"
-        validations:
-          - check: upstream_traffic
-            description: "Partner fetched the licensor's brand.json for licensed_out reciprocation"
-            endpoint_pattern: "GET *non-reciprocating-licensor.example/.well-known/brand.json"
-            min_count: 1
-            identifier_paths:
-              - "claim.mark"
-
-      - id: assert_trademark_trust_not_extended
-        title: "Partner did NOT accept the unverified licensee posture"
-        narrative: |
-          With no reciprocating `licensed_out` from the licensor, the
-          partner MUST NOT have propagated licensee posture into
-          downstream surfaces (creative-clearance auto-approval, mark
-          auto-binding, etc.).
-        task: verify_brand_claim
-        schema_ref: "brand/verify-brand-claim-request.json"
-        response_schema_ref: "brand/verify-brand-claim-response.json"
-        doc_ref: "/brand-protocol/tasks/verify_brand_claim"
-        requires_contract: single_side_trust_runner
-        stateful: false
-        expected: |
-          Zero recorded calls match the licensee-trust-extension
-          endpoint patterns.
-        sample_request:
-          claim_type: trademark
-          claim:
-            mark: "FABRICATED MARK"
-          context:
-            correlation_id: "brand--single_side_trust--trademark--no_trust"
-        validations:
-          - check: upstream_traffic
-            description: "No creative-clearance auto-approval fired on the unverified licensee posture"
-            endpoint_pattern: "POST */creative/auto-clearance"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No trademark auto-binding fired"
-            endpoint_pattern: "POST */trademarks/auto-bind"
-            min_count: 0
-          - check: upstream_traffic
-            description: "No governance/licensing auto-extension fired"
-            endpoint_pattern: "POST */governance/licensing/auto-extend"
-            min_count: 0