test(e2e): migrate test-hermes-sandbox-secret-boundary.sh to vitest

.github/workflows/e2e-vitest-scenarios.yaml

@@ -639,6 +639,50 @@ jobs:
           if-no-files-found: ignore
           retention-days: 14
 
+  hermes-sandbox-secret-boundary-vitest:
+    needs: generate-matrix
+    if: ${{ (inputs.jobs == '' && inputs.scenarios == '') || contains(format(',{0},', inputs.jobs), ',hermes-sandbox-secret-boundary-vitest,') || contains(format(',{0},', inputs.scenarios), ',hermes-sandbox-secret-boundary,') }}
+    runs-on: ubuntu-latest
+    timeout-minutes: 60
+    env:
+      FREE_STANDING_VITEST_JOB: "1"
+      FREE_STANDING_SCENARIO_ID: "hermes-sandbox-secret-boundary"
+      E2E_ARTIFACT_DIR: ${{ github.workspace }}/e2e-artifacts/vitest/hermes-sandbox-secret-boundary
+      NEMOCLAW_RUN_E2E_SCENARIOS: "1"
+    steps:
+      - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3
+        with:
+          persist-credentials: false
+
+      - name: Set up Node
+        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.0.0
+        with:
+          node-version: 22
+          cache: npm
+
+      - name: Install root dependencies
+        run: npm ci --ignore-scripts
+
+      - name: Run Hermes sandbox secret-boundary live test
+        # Migrated from test/e2e/test-hermes-sandbox-secret-boundary.sh. This
+        # builds the real Hermes images unless prebuilt NEMOCLAW_HERMES_* image
+        # env vars are supplied, then probes image and startup secret boundaries.
+        run: |
+          set -euo pipefail
+          npx vitest run --project e2e-scenarios-live \
+            test/e2e-scenario/live/hermes-sandbox-secret-boundary.test.ts \
+            --silent=false --reporter=default
+
+      - name: Upload Hermes sandbox secret-boundary artifacts
+        if: always()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: e2e-vitest-scenarios-hermes-sandbox-secret-boundary
+          path: e2e-artifacts/vitest/hermes-sandbox-secret-boundary/
+          include-hidden-files: false
+          if-no-files-found: ignore
+          retention-days: 14
+
   inference-routing-vitest:
     needs: generate-matrix
     if: ${{ (inputs.jobs == '' && inputs.scenarios == '') || contains(format(',{0},', inputs.jobs), ',inference-routing-vitest,') || contains(format(',{0},', inputs.scenarios), ',inference-routing,') }}
@@ -5326,6 +5370,7 @@ jobs:
         hermes-slack-vitest,
         hermes-discord-vitest,
         hermes-root-entrypoint-smoke-vitest,
+        hermes-sandbox-secret-boundary-vitest,
         network-policy-vitest,
         common-egress-agent-vitest,
         shields-config-vitest,

test/e2e-scenario/fixtures/docker-probe.ts

@@ -29,6 +29,13 @@ export type DockerProbeRunner = (
   options: SpawnSyncOptionsWithStringEncoding,
 ) => SpawnSyncReturns<string>;
 
+type DockerProbeRunOptions = {
+  artifactName: string;
+  timeoutMs?: number;
+  artifactRedactionValues?: string[];
+  returnRaw?: boolean;
+};
+
 const DOCKER_ENV_ALLOWLIST = [
   "DOCKER_HOST",
   "DOCKER_CONTEXT",
@@ -98,7 +105,7 @@ export class DockerProbe {
 
   async run(
     args: string[],
-    options: { artifactName: string; timeoutMs?: number } = { artifactName: "docker" },
+    options: DockerProbeRunOptions = { artifactName: "docker" },
   ): Promise<DockerCommandResult> {
     fs.mkdirSync(this.dockerConfigDir, { recursive: true });
     const command = ["docker", ...args];
@@ -109,30 +116,32 @@ export class DockerProbe {
       maxBuffer: 10 * 1024 * 1024,
       timeout: options.timeoutMs ?? 30_000,
     });
-    const commandResult = redactDockerProbeResult(
-      {
-        command,
-        exitCode: result.status,
-        signal: result.signal,
-        stdout: result.stdout ?? "",
-        stderr: result.stderr ?? "",
-        error: result.error instanceof Error ? result.error.message : undefined,
-      },
-      this.redact,
+    const rawCommandResult = {
+      command,
+      exitCode: result.status,
+      signal: result.signal,
+      stdout: result.stdout ?? "",
+      stderr: result.stderr ?? "",
+      error: result.error instanceof Error ? result.error.message : undefined,
+    };
+    const commandResult = redactDockerProbeResult(rawCommandResult, (text) =>
+      this.redact(text, options.artifactRedactionValues ?? []),
     );
     const artifactBase = `docker/${String(++this.sequence).padStart(3, "0")}-${safeName(
       options.artifactName,
     )}`;
     await this.artifacts.writeText(`${artifactBase}.stdout.txt`, commandResult.stdout);
     await this.artifacts.writeText(`${artifactBase}.stderr.txt`, commandResult.stderr);
     await this.artifacts.writeJson(`${artifactBase}.result.json`, commandResult);
-    return commandResult;
+    return options.returnRaw === true ? rawCommandResult : commandResult;
   }
 
-  async expect(
-    args: string[],
-    options: { artifactName: string; timeoutMs?: number },
-  ): Promise<DockerCommandResult> {
+  async expect(args: string[], options: DockerProbeRunOptions): Promise<DockerCommandResult> {
+    if (options.returnRaw === true) {
+      throw new Error(
+        "DockerProbe.expect cannot return raw Docker output; use run(..., { returnRaw: true }) only for explicit leak assertions that never log the raw result.",
+      );
+    }
     const result = await this.run(args, options);
     if (result.exitCode !== 0) {
       throw new Error(resultText(result));