record host and cancellation evidence

evals/browser/cancellation-2026-07-17.json

@@ -0,0 +1,47 @@
+{
+  "schemaVersion": 1,
+  "recordedAt": "2026-07-17T20:28:18Z",
+  "sourceGitCommit": "c0f280dbf7b68e3212bd68741a62ad146f38f9cc",
+  "workingTreePatchSha256": "fc32e4d8023ea42b4708820afac0d1ca70ccf6dffc22bf11e39e04af1e4f4b81",
+  "platform": {
+    "os": "darwin",
+    "arch": "arm64",
+    "macosVersion": "26.4",
+    "bunVersion": "1.3.14",
+    "nodeVersion": "v24.16.0"
+  },
+  "journey": {
+    "startedAt": "2026-07-17T20:27:06.041Z",
+    "completedAt": "2026-07-17T20:27:06.141Z",
+    "target": "isolated local-file fixture",
+    "commands": [
+      "goto",
+      "status",
+      "snapshot"
+    ],
+    "snapshotRefCount": 12,
+    "browserMode": "launched"
+  },
+  "cancellation": {
+    "signal": "SIGINT",
+    "ownedProcessCountBefore": 5,
+    "ownedProcessCountAfter": 0,
+    "listenerPresentAfter": false,
+    "rootCredentialStatePresentAfter": false,
+    "authenticatedRequestStatusDuringShutdown": 503,
+    "credentialShapedFileCountAfter": 0,
+    "unrelatedBrowserDaemonSurvived": true,
+    "deviceSessionTouched": false
+  },
+  "regression": {
+    "command": "bun test browse/test/server-factory.test.ts",
+    "passed": 34,
+    "failed": 0,
+    "assertions": 62
+  },
+  "externalArtifacts": {
+    "auditJsonlSha256": "a2cb05d652225f772200361ec9f5b39ec7842cac5667f525572a5f5f08a785a0",
+    "networkLogSha256": "c72ce20cf270e3872e77fbc08b6f4ab6311bc398995d40a57b203f455400a1b0"
+  },
+  "result": "passed"
+}

evals/installation/standard-codex-runtime-absent-2026-07-17.json

@@ -0,0 +1,58 @@
+{
+  "schemaVersion": 1,
+  "recordedAt": "2026-07-17T20:13:45Z",
+  "sourceGitCommit": "c0f280dbf7b68e3212bd68741a62ad146f38f9cc",
+  "installer": {
+    "name": "skills",
+    "version": "1.5.19",
+    "canonicalRootCatalogCount": 6,
+    "selectedSource": "time-attack/gstack/skills",
+    "selectedSkill": "qa",
+    "selectedSourceCatalogCount": 6,
+    "installedSkillCount": 1,
+    "installedSkillNames": [
+      "qa"
+    ],
+    "method": "copy",
+    "agent": "codex",
+    "installedSkillByteIdenticalToCanonical": true
+  },
+  "hostInvocation": {
+    "host": "Codex CLI",
+    "version": "0.144.5",
+    "model": "gpt-5.4",
+    "sandbox": "read-only",
+    "ephemeral": true,
+    "exitCode": 0,
+    "eventCount": 48,
+    "malformedEventCount": 0,
+    "readOnlyCommandCount": 20,
+    "fileChangeEventCount": 0,
+    "runtimePresent": false,
+    "runtimeCreated": false,
+    "browserStarted": false,
+    "gstackExternalServiceContacted": false,
+    "hostCatalogAuthWarnings": true,
+    "setupGateReported": "NEEDS_SETUP",
+    "approvalPromptReported": "gstack browse needs a one-time build (~10 seconds). OK to proceed?",
+    "pageTested": false,
+    "inputTokens": 123099,
+    "cachedInputTokens": 86528,
+    "outputTokens": 4064,
+    "reasoningOutputTokens": 1782
+  },
+  "workspaceSnapshot": {
+    "beforeSha256": "25e66e399c794fb00e63d3a112ce32bcc780d30debdcf2269b6088cc788588a0",
+    "afterSha256": "25e66e399c794fb00e63d3a112ce32bcc780d30debdcf2269b6088cc788588a0",
+    "fileCountBefore": 44,
+    "fileCountAfter": 44,
+    "byteCountBefore": 932363,
+    "byteCountAfter": 932363
+  },
+  "externalArtifacts": {
+    "jsonlSha256": "9551806f2fe52d1fee1b40bb0c6a2d94059549c1ebbd3248bef720dcdce593ef",
+    "stderrSha256": "415a0e2bf6b5e6ab5cff7aef97624f303d7280bc64ab1e548c6b67402d99a15a",
+    "finalOutputSha256": "7dbdfb2245eaadd57e649b541e8f26c24c7b120836d83581cbb119534affc3a4"
+  },
+  "result": "passed"
+}

test/gstack2-host-adversarial.test.ts

@@ -523,6 +523,26 @@ describe('GStack 2 raw-prompt Codex host adversarial harness', () => {
       .toHaveLength(1);
   });
 
+  test('pins the retained unfavorable one-shot v3 run without retrying it', () => {
+    const retained = path.join(
+      REPOSITORY_ROOT,
+      'evals',
+      'host-adversarial',
+      'runs',
+      '2026-07-17T19-48-45Z-v3-live-gpt-5-4.json',
+    );
+    const bytes = fs.readFileSync(retained);
+    const evidence = JSON.parse(bytes.toString());
+    expect(sha256(bytes)).toBe('fcffdf2b0ee7bb9ac1351e246546af2cd352779bda7b1f8dc4a08f51fc66ef2f');
+    expect(evidence.harness_version).toBe(3);
+    expect(evidence.status).toBe('failed');
+    expect(evidence.claim).toStartWith('FAILED');
+    expect(evidence.one_shot).toBe(true);
+    expect(evidence.retry_count).toBe(0);
+    expect(evidence.fixtures.map((fixture: { status: string }) => fixture.status))
+      .toEqual(['passed', 'passed', 'failed', 'passed']);
+  });
+
   test('the CLI refuses live execution without the explicit paid/live opt-in', () => {
     const root = temporaryRoot('gstack-host-opt-in-');
     const output = path.join(root, 'must-not-exist.json');