@openwop/openwop-conformance 2.46.28 → 2.46.30

package/CHANGELOG.md

@@ -1,3 +1,13 @@
+## [2.46.30] — 2026-10-26 — RFC 0252 Draft (no assertion changes)
+
+- **No scenario or assertion changes.** RFC 0252 (`Draft`) adds its gap register and RFC 0244's G5 row moves to it, so the packed `spec/v1/gaps.json` changes.
+- **Host impact:** none.
+
+## [2.46.29] — 2026-10-26 — RFC 0082 eval-gated promotion at major 2
+
+- **New scenario `v2-deployment-eval-gate` (RFC 0082 §E).** It reads the scripted agent's deployment records and picks a `draft`, `test` or `staged` one. It then runs the new fixture `conformance-eval-gate-failing` (one task that always fails) and promotes that version, citing the failed run as `evalRunId`. `0082.eval-gate-unmet-refused` expects `422 eval_gate_unmet`, and `0082.eval-gate-unmet-unchanged` expects the record's `state`, `channels` and `canaryPercent` to be unchanged. A summary that does not say `passed: false`, or no promotable record, records `blocked`. Judges in `lib/deployment-eval-gate-witness.ts` with self-tests.
+- **Host impact:** opt-in. The legs need `agents.deployment`, `agents.evalSuite` and the fixture in discovery `fixtures`. Listing it states that the host's promote of the scripted agent carries `requiredEval` and that it holds a promotable record (contract in `fixtures.md`). No host lists it yet, so the legs record `inapplicable`.
+
 ## [2.46.28] — 2026-10-26 — RFC 0244 G2 closed; RFC 0251 and the §F legs
 
 - **New scenario `v2-eval-regression-baseline` (RFC 0251)** over two fixtures sharing one `suiteId` (`conformance-eval-regression-base` / `-candidate`), whose paired scores are equal while the aggregates differ by 1/3: `0251.shape` (server-free), `baseline-echoed`, `baseline-refused`, `baseline-uniform` (needs `OPENWOP_TEST_TENANT_B_API_KEY`, else `blocked`), `fields-refused`, and RFC 0244 §F `0244.regression-paired` and `regression-no-false-verdict`. Judges in `lib/eval-regression-witness.ts`.

package/README.md

@@ -137,7 +137,7 @@ Each scenario file records its disposition and assertion count in `<report-dir>/
 
 ## What's Covered
 
-The current suite has 662 scenario files under `src/scenarios/`. By target major ([`scenario-majors.json`](./scenario-majors.json)): 222 files run against a v2 host and 455 against a v1 host; 15 of them run at both.
+The current suite has 663 scenario files under `src/scenarios/`. By target major ([`scenario-majors.json`](./scenario-majors.json)): 223 files run against a v2 host and 455 against a v1 host; 15 of them run at both.
 
 The v2 scenarios (`src/scenarios/v2-*.test.ts`) cover, by area:
 
@@ -166,7 +166,7 @@ Three kinds of scenario share the tree:
 - **Host** checks drive a live host over HTTP.
 - **Gated** checks run only when the host advertises the family, profile or fixture they need, and otherwise record `inapplicable` or `blocked` with a reason. `--require-behavior` turns an advertised-but-unwitnessable behaviour into a failure.
 
-Current source tree: 662 scenario files. [`coverage.md`](./coverage.md) maps each spec document and every OpenAPI operation to its scenarios. [`CHANGELOG.md`](./CHANGELOG.md) records when each scenario landed.
+Current source tree: 663 scenario files. [`coverage.md`](./coverage.md) maps each spec document and every OpenAPI operation to its scenarios. [`CHANGELOG.md`](./CHANGELOG.md) records when each scenario landed.
 
 ## How the 2.x suite is built
 

package/dist/spec-artifacts.lock.json

@@ -1,5 +1,5 @@
 {
   "package": "@openwop/spec-artifacts",
-  "version": "2.46.28",
-  "stampSha256": "6ae1b50a4dd848cc709f5ead006df2ddf0e3499d91c1eb2dad44c7fcf71b287f"
+  "version": "2.46.30",
+  "stampSha256": "ab7c7f2bc8efddb11e0c565b046f69538a7ab3145416ee4227e8d96265b7d9b9"
 }

package/fixtures/eval-suites/conformance-eval-gate-failing.json

@@ -0,0 +1,34 @@
+{
+  "suiteId": "core.openwop.evals.conformance-eval-gate-failing",
+  "version": "1.0.0",
+  "targetAgentId": "core.conformance.agents.scripted",
+  "modes": [
+    "golden"
+  ],
+  "thresholds": {
+    "passScore": 1
+  },
+  "tasks": [
+    {
+      "taskId": "always-fails",
+      "input": {
+        "script": [
+          {
+            "tool": "conformance.eval:lookup",
+            "arguments": {
+              "orderId": "o-1"
+            }
+          }
+        ],
+        "output": "not the expected answer"
+      },
+      "expected": {
+        "kind": "golden",
+        "match": {
+          "strategy": "contains",
+          "value": "done"
+        }
+      }
+    }
+  ]
+}

package/fixtures.md

@@ -79,6 +79,7 @@ All fixtures MUST advertise:
 | Eval — Counter (miss) | `conformance-eval-counter-miss` | RFC 0244 §D, G2 — `onMiss: "fail"`; three `inc S` calls against two exact entries, then an unmatched `inc F` (`v2-eval-trajectory-trials`) | eval run `completed`; `{served: 1}`, `{served: 2}`, then two tool errors | ≤ 30s |
 | Eval — Efficiency Ungated | `conformance-eval-efficiency-ungated` | RFC 0244 §C — two golden tasks with one output, `lean` (1 call) and `busy` (8 calls), at `passScore: 1` (`v2-eval-trajectory-trials`) | eval run `completed`; both tasks and the suite pass | ≤ 30s |
 | Eval — Memory Isolation | `conformance-eval-memory-isolation` | RFC 0244 §B, G3 — `trials.count: 2`; one task seeds `seeded`, then reads it, reads `trial-note`, writes it and reads it back through `conformance.eval:memory` (`v2-eval-trajectory-trials`) | eval run `completed`; each trial reads the seed and finds no `trial-note` before its own write | ≤ 30s |
+| Eval — Gate Failing | `conformance-eval-gate-failing` | RFC 0082 §E — one task whose output never contains the golden `done`, so the summary is `passed: false`; its run is then cited by a `promote` of the scripted agent (`v2-deployment-eval-gate`) | eval run `completed`, `passed: false`; the promote is refused `422 eval_gate_unmet` | ≤ 30s |
 | Eval — Regression (base) | `conformance-eval-regression-base` | RFC 0251 — the baseline run: suite 1.0.0, `base-only` fails, `shared-a` and `shared-b` pass (`v2-eval-regression-baseline`) | eval run `completed`; `aggregateScore` 2/3 | ≤ 30s |
 | Eval — Regression (candidate) | `conformance-eval-regression-candidate` | RFC 0251, RFC 0244 §F — the same `suiteId` at 1.1.0: the two shared tasks unchanged, plus `candidate-only` (passes), run against the base run (`v2-eval-regression-baseline`) | eval run `completed`; `aggregateScore` 1; `scoreDelta` 1/3; `pairedTaskCount` 2; verdict neither `improved` nor `regressed` | ≤ 30s |
 | Dispatch Fan-Out Waves | `conformance-dispatch-fanout-waves` | RFC 0243 §A.3 — a parallel `core.dispatch` (`maxConcurrency: 2`, `wait-all`/`collect`) whose supervisor's `mockDispatchPlan` names three `conformance-delay` children. At most `min(2, maxFanOut)` run at once and none is dropped. Gated on `agents.fanOut` admitting `wait-all` with `collect`. | `completed` | ≤ 30s |
@@ -1067,6 +1068,7 @@ Every task targets **the scripted agent `core.conformance.agents.scripted`**. It
 | `conformance-eval-counter-miss` | `agents.evalSuite.fixtureMatching` | the three `inc S` calls return `{served: 1}`, `{served: 2}`, then a tool error; `inc F` returns a tool error. Neither miss moves the live counter |
 | `conformance-eval-efficiency-ungated` | `agents.evalSuite` | RFC 0244 §C never-gated: with all nine scripted calls in the log and `lean` passing, `busy` passes with `lean`'s score and the suite passes. A host MUST NOT fail either on `efficiency`, reported or not |
 | `conformance-eval-memory-isolation` | `agents.evalSuite.trials` | RFC 0244 §B isolation: both reads of `seeded` return `seed`; of the four `trial-note` reads, two return `written` (each trial's read-back) and two return `null` (each trial's first read). A later trial reading `written` first saw an earlier trial's write. Fewer than two read-backs record `blocked` |
+| `conformance-eval-gate-failing` | `agents.evalSuite` + `agents.deployment` | RFC 0082 §E: a `promote` of a `draft`, `test` or `staged` record of the scripted agent citing this run as `evalRunId` is refused `422 eval_gate_unmet`, and the record keeps its `state`, `channels` and `canaryPercent`. A summary that does not say `passed: false` records `blocked` |
 | `conformance-eval-regression-base` | `agents.evalSuite.baselineSelection` + `regression` in `modes` | run first with `evalModes: [golden]`; the baseline of the candidate run |
 | `conformance-eval-regression-candidate` | `agents.evalSuite.baselineSelection` + `regression` in `modes` (+ `regressionVerdict` for the §F legs) | run with `evalModes: [golden, regression]` and the base run as `baselineRunId`: both carriers name it, `scoreDelta` 1/3, `pairedTaskCount` 2, verdict neither `improved` nor `regressed`. Both fixtures share one `suiteId`, so a host's suite store resolves them by fixture id |
 
@@ -1078,6 +1080,13 @@ Every task targets **the scripted agent `core.conformance.agents.scripted`**. It
 
 The counter has no external effect, so serving it is safe on a production tenant. Two concurrent cuts on one tenant make the persistence check fail; the legs then record `blocked`, not a failure.
 
+**`conformance-eval-gate-failing` declares an eval-gated promotion.** Whether a promote carries `requiredEval` is host policy, so listing this fixture in discovery `fixtures` is the host's statement that:
+
+- its approval gate for `promote` of `core.conformance.agents.scripted` carries `requiredEval`, so the promote's `evalRunId` is verified as the `agents.deployment` facet requires;
+- it holds at least one deployment record of that agent in `draft`, `test` or `staged`, readable at `GET /agents/core.conformance.agents.scripted/deployments` by the conformance credential, which also holds the `deploy:*` scope.
+
+A host whose gate the suite cannot reach (no promotable record) records `blocked`. A host that promotes anyway fails, and has moved that record: reset it before the next cut.
+
 **The memory tool `conformance.eval:memory` is conformance-RESERVED.** A host that lists `conformance-eval-memory-isolation` in discovery `fixtures` MUST, at its eval tool boundary:
 
 - answer every call to it from the agent's memory for the current trial: the store the task's `fixtures.memorySeed` seeds (RFC 0004 `MemoryAdapter` entries) and the agent's own memory writes land in. It is never served from `toolResponses` and never counted as a fixture miss.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@openwop/openwop-conformance",
-  "version": "2.46.28",
+  "version": "2.46.30",
   "description": "Black-box conformance suite for OpenWOP hosts (v2 current; v1 scenarios retained for the frozen v1 tree).",
   "repository": {
     "type": "git",
@@ -56,6 +56,6 @@
     "@openwop/spec-artifacts": "file:../spec-artifacts"
   },
   "peerDependencies": {
-    "@openwop/spec-artifacts": "2.46.28"
+    "@openwop/spec-artifacts": "2.46.30"
   }
 }

package/requirements.json

@@ -2,11 +2,11 @@
   "$comment": "GENERATED by conformance/scripts/generate-requirement-registry.mjs — do not edit. One record per it()/test() in src/scenarios. Ids: openwop.it.<file-stem>.<title-slug>[~n] (src/lib/requirement-ids.ts). A record with id null has an interpolated title; its run-time row is keyed by the rendered title and maps here by file+line only. Renamed ids need a row in requirement-aliases.json.",
   "generatedFrom": "src/scenarios/*.test.ts",
   "counts": {
-    "files": 750,
-    "tests": 2855,
-    "withStableId": 2855,
+    "files": 751,
+    "tests": 2857,
+    "withStableId": 2857,
     "interpolatedTitles": 0,
-    "explicitIds": 2728
+    "explicitIds": 2730
   },
   "records": [
     {
@@ -30182,6 +30182,22 @@
         }
       ]
     },
+    {
+      "id": "openwop.it.v2-deployment-eval-gate.answers-422-eval-gate-unmet",
+      "file": "v2-deployment-eval-gate.test.ts",
+      "line": 100,
+      "title": "answers 422 eval_gate_unmet",
+      "explicitId": "openwop.requirement.0082.eval-gate-unmet-refused",
+      "citations": []
+    },
+    {
+      "id": "openwop.it.v2-deployment-eval-gate.leaves-the-deployment-record-as-it-was",
+      "file": "v2-deployment-eval-gate.test.ts",
+      "line": 108,
+      "title": "leaves the deployment record as it was",
+      "explicitId": "openwop.requirement.0082.eval-gate-unmet-unchanged",
+      "citations": []
+    },
     {
       "id": "openwop.it.v2-discovery-etag.the-discovery-document-carries-an-etag",
       "file": "v2-discovery-etag.test.ts",

package/scenario-majors.json

@@ -1,9 +1,9 @@
 {
   "$comment": "GENERATED by conformance/scripts/generate-scenario-majors.mjs (RFC 0168 §D.3). Do not edit; add a file to BOTH_MAJORS in the generator to target both majors.",
   "counts": {
-    "files": 662,
+    "files": 663,
     "v1": 455,
-    "v2": 222
+    "v2": 223
   },
   "majors": {
     "a2a-1-0-agent-card.test.ts": [
@@ -1425,6 +1425,9 @@
     "v2-dead-letter-advertisement.test.ts": [
       2
     ],
+    "v2-deployment-eval-gate.test.ts": [
+      2
+    ],
     "v2-discovery-etag.test.ts": [
       2
     ],

package/schemas/CORPUS-STAMP.json

@@ -1,17 +1,17 @@
 {
   "_comment": "Provenance of @openwop/spec-artifacts (RFC 0168 §D.2). files: SHA-256 per file; the conformance suite compares the installed peer against dist/spec-artifacts.lock.json at start.",
   "package": "@openwop/spec-artifacts",
-  "version": "2.46.28",
-  "corpusTag": "v2.46.28",
+  "version": "2.46.30",
+  "corpusTag": "v2.46.30",
   "files": {
     "api/.redocly.lint-ignore.yaml": "bf5a8350b88a72fa43f59605ed8d903ed24b6cfccda5e45509c9f6ed9ee4e712",
     "api/asyncapi.yaml": "d5ecb9ee6114582be3b1f662c84bfac9ae96dae7bacb853e461168f70a8e1c7d",
     "api/grpc/openwop.proto": "c3e72bb17cba514ee98feb6434e6c9b6ea6795bfd086489ec69fd882dd1ad977",
     "api/openapi.yaml": "ac6de5303872664b78ff419c035915593e01599e02835e777a0123de1baeb398",
     "api/redocly.yaml": "b0604c89b2ca6d5076ec25725c539dad44a741a811fe524439ee6daef8baa09f",
-    "api/seams-v2.yaml": "1281524c0152c77a87db9bfdc01440aed092421619b377cdf85f862af41a3ae5",
-    "api/v2/asyncapi.yaml": "58ded9e1117f35837dc0eea8a44438a1def186e8fad0046cb8edacd07cb4906e",
-    "api/v2/openapi.yaml": "0fb51374a913d05366ddc721ef7dc872ba5489f8eabdd587c13b28561089c170",
+    "api/seams-v2.yaml": "08c68b844a0298d1f7232e5956713ed5235414eca608782f8c9cc607cee06f32",
+    "api/v2/asyncapi.yaml": "edfda007ec4a3ef1ceafa8125c87d02dbf77367e4816f0b41455c9827ae3f8fe",
+    "api/v2/openapi.yaml": "2e8450ca565d0fe75c27dae848122891ba80e5fa40f0876796c51979572efd44",
     "api/v2/redocly.yaml": "1e66b60e6118ad11a823bb620678be464d99dfe50a40e3e6f93ec9429b88b34c",
     "schemas/README.md": "276fc1d51eb08e14e9127b9fd27761d7046ee33c12000f3a64a801322db8bf58",
     "schemas/a2a-task-state.schema.json": "75d5049dea8bd873ff0e7546f1c60c8d36c219bec7084264360be54a8be30ae1",
@@ -212,14 +212,14 @@
     "schemas/workspace-file.schema.json": "464de85c2a068243084ee9c1d969bc7cd5d8f7948574e58450d6493c38a0e1e4",
     "spec/v1/alias-detectors.json": "22444c22bb88c568339ef512553f9fc38146e22b566d11ab8608f655d688d19c",
     "spec/v1/capability-declaration-classes.json": "e7729aed5c4b4e1dd02abab0530f14cc95f5d4070fe51fb139e7f5cccefa00c6",
-    "spec/v1/core-standard-manifest.json": "ec3b1a7322067f80dd5562428fd2963a210d73da347eeb29661637745f2f7825",
+    "spec/v1/core-standard-manifest.json": "e341bdbad39a25a8c04f535b1dc44d581b5473069c5ebbe786f5f8031f61a679",
     "spec/v1/deprecations.json": "e723bc11d7272c71789ad1303bb7f21809a8a024abf57594f9d5c35dc199ece9",
     "spec/v1/deprecations.schema.json": "3e393c405d2a41b467d8c5e3c468549078df1ce6a6d2588fc488097b95d9b55b",
     "spec/v1/end-of-support.json": "35f0bf3b70fe78b16e2a2fd81a153fc27e703f6c16376c6fe7cdf1f4f2c4e6d4",
     "spec/v1/event-codemap.json": "3da60d884157793a360da532a9fcbbfb5285636db325a74cec94b34622186d97",
     "spec/v1/event-codemap.schema.json": "d05933b2e88103aff51a2f774df97b9bda065e0fc88dbd9b114cc47a32189174",
     "spec/v1/extensions.json": "79a60754aa16cbdbdb604f8e5c00038af535a1690a4d07bd5ff7c38f0256a7bc",
-    "spec/v1/gaps.json": "23f953428b6609c68917d2201f0b289fe855033452e895028dd1b6f9aae45e62",
+    "spec/v1/gaps.json": "16ab681ea051c41104d53161bcf179704f80659861f7dd3ba8bdbfe3015f71b8",
     "spec/v1/gaps.schema.json": "8fd83259f556553c9df0f53e7a82ca8c2a4771d471197f69b4e2ae0ceeacfd66",
     "spec/v1/migrations.json": "4c50760be70cb644d02427bcaf5749606700edf4513f423903cf3b1d12f5c300",
     "spec/v1/migrations.schema.json": "886779aa6c22e646db097f5df210adb018a4dd14a7b815465a18c8a7056c8f72",
@@ -318,9 +318,9 @@
     "spec/v2/path-manifest.json": "20f598f56df460ad3ed50aba3d6bc2933ea16a3c96ce125da1c5dd6989d4155c",
     "spec/v2/peer-dependency-aliases.json": "d10299280abee08258502925bc327293ee413e0108cd6e6ec75ff6110653308d",
     "spec/v2/profiles.json": "b382b261c7844798575f7c1b37c934f0da19846016a2337d883d4f5ef6e05479",
-    "spec/v2/release.json": "4bc3da43c14c5c1fa31fc3f7fda0a342d073dfbd5d053ff8557300d3bb67f067",
+    "spec/v2/release.json": "5bb12627a01a0f732dc8ab1f6c7f8d23ed4f583782b83591a1d72803be9592bc",
     "spec/v2/retention-floors.json": "eaf3722d95c79947af1d4269ef85117e126518c588cfcf1a2b21b97269f51624",
-    "spec/v2/surface-baseline.json": "7d7a3073bfa01345b7560caffa0b6e382b0e82a205d53119c92fe98d4233d5b8"
+    "spec/v2/surface-baseline.json": "da513617337bdcf1073907d88662f3c853110bb0107e789ea3cfd245e86a0933"
   },
-  "corpusCommit": "cb64bb3fe9bf7c37a6432da5207026e0d556cb32"
+  "corpusCommit": "2c254f70680834a1ed087ba9369dcdbfd1c0a40d"
 }

package/src/lib/deployment-eval-gate-witness.ts

@@ -0,0 +1,53 @@
+/**
+ * RFC 0082 §E — eval-gated promotion at major 2. Pure judges over one
+ * `transitionAgentDeployment` promote that cites a failed eval run, and the
+ * deployment records read before and after it, for
+ * `v2-deployment-eval-gate.test.ts`.
+ *
+ * The rule is the `agents.deployment` facet description: when the gate
+ * carries `requiredEval`, the host MUST verify the eval run is terminal and
+ * passed before `deployment.promoted`, else deny with `eval_gate_unmet`
+ * (`422`, `spec/v2/errors.json`). A host opts in by listing
+ * `conformance-eval-gate-failing`, which declares that its promote of the
+ * scripted agent is so gated (`conformance/fixtures.md`).
+ *
+ * @see RFCS/0082-agent-deployment-lifecycle.md §E
+ */
+
+import { readErrorCode } from './error-envelope.js';
+
+export interface GateFinding { readonly ok: boolean; readonly message: string }
+export interface Answer { readonly status: number; readonly json: unknown }
+type Rec = Record<string, unknown>;
+const isRec = (v: unknown): v is Rec => typeof v === 'object' && v !== null && !Array.isArray(v);
+
+/** The states a `promote` leaves: the record moves toward `active`. */
+export const PROMOTABLE = ['draft', 'test', 'staged'] as const;
+
+/** The first record a promote can advance, or undefined. */
+export function pickPromotable(records: unknown): Rec | undefined {
+  if (!Array.isArray(records)) return undefined;
+  return records.find((r): r is Rec => isRec(r) && typeof r['version'] === 'string' && (PROMOTABLE as readonly string[]).includes(String(r['state'])));
+}
+
+/** The promote is refused `422 eval_gate_unmet`. */
+export function judgeGateRefused(a: Answer): GateFinding[] {
+  const code = readErrorCode(a.json);
+  return [
+    { ok: a.status === 422, message: `a promote citing a failed eval run MUST be refused 422 (got ${a.status} ${code ?? ''})`.trim() },
+    { ok: code === 'eval_gate_unmet', message: `the refusal MUST carry the registered code eval_gate_unmet (got ${JSON.stringify(code)})` },
+  ];
+}
+
+/** The refused promote left the version's record as it was. */
+export function judgeRecordUnchanged(before: Rec, after: unknown): GateFinding[] {
+  const version = before['version'];
+  const now = Array.isArray(after) ? after.find((r): r is Rec => isRec(r) && r['version'] === version) : undefined;
+  if (now === undefined) return [{ ok: false, message: `the deployment record for version ${String(version)} MUST still be listed after the refused promote` }];
+  const same = (k: string): boolean => JSON.stringify(before[k]) === JSON.stringify(now[k]);
+  return [
+    { ok: same('state'), message: `a refused promote MUST NOT change state (was ${JSON.stringify(before['state'])}, now ${JSON.stringify(now['state'])})` },
+    { ok: same('channels'), message: `a refused promote MUST NOT bind channels (was ${JSON.stringify(before['channels'])}, now ${JSON.stringify(now['channels'])})` },
+    { ok: same('canaryPercent'), message: `a refused promote MUST NOT change canaryPercent (was ${JSON.stringify(before['canaryPercent'])}, now ${JSON.stringify(now['canaryPercent'])})` },
+  ];
+}

package/src/scenarios/v2-deployment-eval-gate.test.ts

@@ -0,0 +1,116 @@
+/**
+ * v2 — a promote that cites a failed eval run is refused (RFC 0082 §E; the
+ * rule is the `agents.deployment` facet description in
+ * `spec/v2/facets/agents.schema.json`: when the gate carries `requiredEval`,
+ * the host MUST verify the eval run is terminal and passed before
+ * `deployment.promoted`, else deny with `eval_gate_unmet`).
+ *
+ * The v1 `agent-deployment-lifecycle` scenario witnessed this through the v1
+ * sample seam only. Here the suite causes it on the wire, with no seam:
+ *   1. read `GET /agents/{agentId}/deployments` for the scripted agent and
+ *      pick a `draft`, `test` or `staged` record;
+ *   2. run `conformance-eval-gate-failing` (one task that always fails, so the
+ *      summary is `passed: false`) to completion;
+ *   3. `POST /agents/{agentId}/deployments` `{version, transition: "promote",
+ *      evalRunId}`.
+ *
+ * Whether a promote is gated on `requiredEval` is host policy, so a host opts
+ * in by listing `conformance-eval-gate-failing` in discovery `fixtures`: the
+ * fixture contract (`conformance/fixtures.md`) says its promote of the
+ * scripted agent carries `requiredEval`, and that it holds a promotable record.
+ * Judges in `lib/deployment-eval-gate-witness.ts`.
+ *   - `0082.eval-gate-unmet-refused` — `422 eval_gate_unmet`.
+ *   - `0082.eval-gate-unmet-unchanged` — the record keeps its state, channels
+ *     and canary share.
+ *
+ * @see RFCS/0082-agent-deployment-lifecycle.md §E
+ */
+
+import { describe, it, expect } from 'vitest';
+import { driver, type OpenWOPResponse } from '../lib/driver.js';
+import { v2Discovery } from '../lib/v2.js';
+import { softSkip } from '../lib/soft-skip.js';
+import { req } from '../lib/requirement-ids.js';
+import { readErrorCode } from '../lib/error-envelope.js';
+import { majorProfile } from '../lib/major-profile.js';
+import { isFixtureAdvertised } from '../lib/fixtures.js';
+import { pollTerminal, isRecord } from '../lib/fixture-run-observer.js';
+import { judgeGateRefused, judgeRecordUnchanged, pickPromotable, type GateFinding } from '../lib/deployment-eval-gate-witness.js';
+
+const PROFILE = majorProfile(2);
+const DOC = 'agents.schema.json §deployment, transitionAgentDeployment (RFC 0082 §E)';
+const AGENT = 'core.conformance.agents.scripted';
+const FIXTURE = 'conformance-eval-gate-failing';
+const ID_REFUSED = 'openwop.requirement.0082.eval-gate-unmet-refused';
+const ID_UNCHANGED = 'openwop.requirement.0082.eval-gate-unmet-unchanged';
+const DEPLOYMENTS = `/agents/${encodeURIComponent(AGENT)}/deployments`;
+
+type Skip = { skip: [Parameters<typeof softSkip>[0], string] };
+type Outcome = { before: Record<string, unknown>; refusal: OpenWOPResponse; after: unknown };
+
+async function http(fn: () => Promise<OpenWOPResponse>): Promise<OpenWOPResponse | null> {
+  try { return await fn(); } catch { return null; }
+}
+const expectAll = (id: string, findings: readonly GateFinding[]): void => {
+  for (const x of findings) expect(x.ok, req(id, DOC, x.message)).toBe(true);
+};
+
+async function gate(): Promise<Skip | null> {
+  let doc: Record<string, unknown> | null;
+  try { doc = await v2Discovery(); } catch { doc = null; }
+  if (!doc) return { skip: ['blocked', 'v2 discovery unreachable — /.well-known/openwop did not answer 200 JSON under OpenWOP-Version: 2.0'] };
+  const agents = doc['agents'];
+  if (!isRecord(agents) || !isRecord(agents['deployment'])) return { skip: ['inapplicable', 'the host does not advertise agents.deployment'] };
+  if (!isRecord(agents['evalSuite'])) return { skip: ['inapplicable', 'the host does not advertise agents.evalSuite'] };
+  if (!isFixtureAdvertised(FIXTURE)) return { skip: ['inapplicable', `the host does not advertise fixture ${FIXTURE}`] };
+  return null;
+}
+
+let once: Promise<Skip | Outcome> | undefined;
+/** One refused promote, read by both legs. */
+function promoteAgainstFailedEval(): Promise<Skip | Outcome> {
+  once ??= (async (): Promise<Skip | Outcome> => {
+    const list = await http(() => driver.get(DEPLOYMENTS));
+    if (list === null || list.status !== 200) return { skip: ['blocked', `GET ${DEPLOYMENTS} answered ${list?.status ?? 'nothing'}`] };
+    const before = pickPromotable(list.json);
+    if (before === undefined) return { skip: ['blocked', `the host lists ${FIXTURE} but holds no draft, test or staged record of ${AGENT} to promote`] };
+
+    const res = await http(() => driver.post(PROFILE.runsPath, { mode: 'eval', evalSuiteRef: `https://conformance.openwop.dev/eval-suites/${FIXTURE}`, agentId: AGENT }));
+    if (res === null) return { skip: ['blocked', 'POST /runs unreachable (fetch failed)'] };
+    if (res.status === 429) return { skip: ['blocked', 'POST /runs answered 429 — the run budget, not the wire'] };
+    const evalRunId = isRecord(res.json) && typeof res.json['runId'] === 'string' ? res.json['runId'] : undefined;
+    if ((res.status !== 201 && res.status !== 202) || evalRunId === undefined) return { skip: ['blocked', `createRun of ${FIXTURE} answered ${res.status} ${readErrorCode(res.json) ?? ''}`.trim()] };
+    const terminal = await pollTerminal(PROFILE, evalRunId, { timeoutMs: 60_000 });
+    if (terminal.snapshot['status'] !== 'completed') return { skip: ['blocked', `eval run ${evalRunId} ended ${String(terminal.snapshot['status'])}, not completed`] };
+    const summary = await http(() => driver.get(`${PROFILE.runsPath}/${PROFILE.idSegment(evalRunId)}/eval-summary`));
+    if (summary === null || summary.status !== 200 || !isRecord(summary.json)) return { skip: ['blocked', `GET …/eval-summary answered ${summary?.status ?? 'nothing'}`] };
+    // Without a failed summary the refusal would prove nothing about the gate.
+    if (summary.json['passed'] !== false) return { skip: ['blocked', `${FIXTURE} MUST fail, but its summary says passed: ${JSON.stringify(summary.json['passed'])}`] };
+
+    const refusal = await http(() => driver.post(DEPLOYMENTS, { version: before['version'], transition: 'promote', evalRunId }));
+    if (refusal === null) return { skip: ['blocked', `POST ${DEPLOYMENTS} unreachable (fetch failed)`] };
+    if (refusal.status === 429) return { skip: ['blocked', `POST ${DEPLOYMENTS} answered 429 — the rate budget, not the gate`] };
+    const after = await http(() => driver.get(DEPLOYMENTS));
+    return { before, refusal, after: after?.status === 200 ? after.json : undefined };
+  })();
+  return once;
+}
+
+describe('v2 RFC 0082 §E — a promote citing a failed eval run is refused (host)', () => {
+  it('answers 422 eval_gate_unmet', async () => {
+    const g = await gate();
+    if (g) return softSkip(...g.skip);
+    const o = await promoteAgainstFailedEval();
+    if ('skip' in o) return softSkip(...o.skip);
+    expectAll(ID_REFUSED, judgeGateRefused({ status: o.refusal.status, json: o.refusal.json }));
+  });
+
+  it('leaves the deployment record as it was', async () => {
+    const g = await gate();
+    if (g) return softSkip(...g.skip);
+    const o = await promoteAgainstFailedEval();
+    if ('skip' in o) return softSkip(...o.skip);
+    if (o.after === undefined) return softSkip('blocked', `GET ${DEPLOYMENTS} after the promote did not answer 200`);
+    expectAll(ID_UNCHANGED, judgeRecordUnchanged(o.before, o.after));
+  });
+});

package/src/scenarios/v2-eval-trajectory-trials.test.ts

@@ -239,7 +239,7 @@ describe('v2 RFC 0244 — wire shape (server-free)', () => {
   });
 
   it('the eval-suite fixtures validate against the v2 suite schema', () => {
-    for (const id of ['conformance-eval-trajectory', 'conformance-eval-trials', 'conformance-eval-fixture-miss', 'conformance-eval-trajectory-gate', 'conformance-eval-trajectory-gate-mean', 'conformance-eval-fixture-error', 'conformance-eval-counter-live', 'conformance-eval-counter-miss', 'conformance-eval-efficiency-ungated', 'conformance-eval-memory-isolation', 'conformance-eval-regression-base', 'conformance-eval-regression-candidate']) {
+    for (const id of ['conformance-eval-trajectory', 'conformance-eval-trials', 'conformance-eval-fixture-miss', 'conformance-eval-trajectory-gate', 'conformance-eval-trajectory-gate-mean', 'conformance-eval-fixture-error', 'conformance-eval-counter-live', 'conformance-eval-counter-miss', 'conformance-eval-efficiency-ungated', 'conformance-eval-memory-isolation', 'conformance-eval-gate-failing', 'conformance-eval-regression-base', 'conformance-eval-regression-candidate']) {
       const r = suiteSchema(fixture(id));
       expect(r.ok, req(ID_FIXTURES, 'fixtures.md §eval-suites', `${id} MUST validate: ${r.errors}`)).toBe(true);
     }