test(e2e): wire sandbox operations into Vitest

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

@@ -4150,6 +4150,117 @@ jobs:
           docker logout docker.io || true
           rm -rf "${DOCKER_CONFIG}"
 
+  sandbox-operations-vitest:
+    needs: generate-matrix
+    if: ${{ (inputs.jobs == '' && inputs.scenarios == '') || contains(format(',{0},', inputs.jobs), ',sandbox-operations-vitest,') || contains(format(',{0},', inputs.scenarios), ',sandbox-operations,') }}
+    runs-on: ubuntu-latest
+    # The live test receives 45 minutes for two onboards plus process/gateway
+    # recovery. The remaining 15 minutes cover checkout, build, OpenShell setup,
+    # artifact upload, and unconditional credential/resource cleanup.
+    timeout-minutes: 60
+    env:
+      FREE_STANDING_VITEST_JOB: "1"
+      FREE_STANDING_SCENARIO_ID: "sandbox-operations"
+      E2E_ARTIFACT_DIR: ${{ github.workspace }}/e2e-artifacts/vitest/sandbox-operations
+      NEMOCLAW_CLI_BIN: ${{ github.workspace }}/bin/nemoclaw.js
+      NEMOCLAW_RUN_E2E_SCENARIOS: "1"
+      NEMOCLAW_E2E_USE_HOSTED_INFERENCE: "1"
+      NEMOCLAW_NON_INTERACTIVE: "1"
+      NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE: "1"
+      # Open permits the scenario's inference/log probes; TC-SBX-11 separately
+      # proves that sandbox-to-sandbox network isolation remains enforced.
+      NEMOCLAW_POLICY_TIER: "open"
+      OPENSHELL_GATEWAY: "nemoclaw"
+    steps:
+      - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3
+        with:
+          persist-credentials: false
+
+      - name: Set up Node
+        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.0.0
+        with:
+          node-version: 22
+          cache: npm
+
+      - name: Install root dependencies
+        run: npm ci --ignore-scripts
+
+      - name: Build CLI
+        run: npm run build:cli
+
+      - name: Verify CLI launcher
+        run: |
+          test -x "${NEMOCLAW_CLI_BIN}"
+          "${NEMOCLAW_CLI_BIN}" --version
+
+      - name: Install OpenShell CLI
+        run: |
+          env -u DOCKER_CONFIG \
+            -u DOCKERHUB_USERNAME \
+            -u DOCKERHUB_TOKEN \
+            -u NVIDIA_API_KEY \
+            -u NVIDIA_INFERENCE_API_KEY \
+            -u GITHUB_TOKEN \
+            bash scripts/install-openshell.sh
+
+      - name: Configure isolated Docker auth directory
+        run: echo "DOCKER_CONFIG=${RUNNER_TEMP}/docker-config-sandbox-operations" >> "$GITHUB_ENV"
+
+      - name: Authenticate to Docker Hub
+        env:
+          DOCKERHUB_USERNAME: ${{ secrets.DOCKERHUB_USERNAME }}
+          DOCKERHUB_TOKEN: ${{ secrets.DOCKERHUB_TOKEN }}
+        shell: bash
+        run: |
+          set -euo pipefail
+          if [[ -z "${DOCKERHUB_USERNAME}" || -z "${DOCKERHUB_TOKEN}" ]]; then
+            echo "::notice::Docker Hub credentials not configured; continuing with anonymous pulls."
+            exit 0
+          fi
+          mkdir -p "${DOCKER_CONFIG}"
+          chmod 700 "${DOCKER_CONFIG}"
+          login_succeeded=0
+          for attempt in 1 2 3; do
+            if echo "${DOCKERHUB_TOKEN}" | timeout 30s docker login docker.io --username "${DOCKERHUB_USERNAME}" --password-stdin; then
+              login_succeeded=1
+              break
+            fi
+            if [[ "$attempt" -lt 3 ]]; then
+              echo "::warning::Docker Hub login attempt ${attempt} failed; retrying."
+              sleep 5
+            fi
+          done
+          if [[ "$login_succeeded" -ne 1 ]]; then
+            echo "::warning::Docker Hub login failed after 3 attempts; continuing with anonymous pulls."
+          fi
+
+      - name: Run sandbox operations live test
+        env:
+          NVIDIA_INFERENCE_API_KEY: ${{ secrets.NVIDIA_INFERENCE_API_KEY }}
+        run: |
+          set -euo pipefail
+          export PATH="$HOME/.local/bin:$HOME/.npm-global/bin:$PATH"
+          npx vitest run --project e2e-scenarios-live \
+            test/e2e-scenario/live/sandbox-operations.test.ts \
+            --silent=false --reporter=default
+
+      - name: Upload sandbox operations artifacts
+        if: always()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: e2e-vitest-scenarios-sandbox-operations
+          path: e2e-artifacts/vitest/sandbox-operations/
+          include-hidden-files: false
+          if-no-files-found: ignore
+          retention-days: 14
+
+      - name: Clean up Docker auth
+        if: always()
+        run: |
+          set -euo pipefail
+          docker logout docker.io || true
+          rm -rf "${DOCKER_CONFIG}"
+
   sandbox-survival-vitest:
     needs: generate-matrix
     if: ${{ (inputs.jobs == '' && inputs.scenarios == '') || contains(format(',{0},', inputs.jobs), ',sandbox-survival-vitest,') || contains(format(',{0},', inputs.scenarios), ',sandbox-survival,') }}
@@ -5785,6 +5896,7 @@ jobs:
         issue-4462-scope-upgrade-approval-vitest,
         onboard-resume-vitest,
         model-router-provider-routed-inference-vitest,
+        sandbox-operations-vitest,
         sandbox-survival-vitest,
         diagnostics-vitest,
         snapshot-commands-vitest,

test/e2e-scenario/fixtures/clients/host.ts

@@ -7,16 +7,21 @@ import { trustedShellCommand } from "../shell-probe.ts";
 import {
   artifactLabel,
   assertExitZero,
+  type CommandRunner,
   outputContainsSandbox,
   resultText,
-  type CommandRunner,
 } from "./command.ts";
 
 export interface HostClientOptions {
   cliPath?: string;
   cwd?: string;
 }
 
+const GATEWAY_ALREADY_ABSENT =
+  /gateway[^\n]*(?:does not exist|not found)|No (?:active )?gateway|No gateway metadata found/i;
+const GATEWAY_REMOVE_UNSUPPORTED =
+  /unrecognized subcommand ['"]remove['"]|unknown command ['"]remove['"]/i;
+
 export class HostCliClient {
   private readonly runner: CommandRunner;
   private readonly cliPath: string;
@@ -122,6 +127,30 @@ export class HostCliClient {
     assertExitZero(result, `cleanup destroy sandbox ${sandboxName}`);
   }
 
+  async cleanupGatewayRegistration(
+    gatewayName: string,
+    options: ShellProbeRunOptions = {},
+  ): Promise<void> {
+    const artifactName = options.artifactName ?? `cleanup-gateway-${artifactLabel(gatewayName)}`;
+    const remove = await this.command("openshell", ["gateway", "remove", gatewayName], {
+      ...options,
+      artifactName: `${artifactName}-remove`,
+    });
+    if (remove.exitCode === 0 || GATEWAY_ALREADY_ABSENT.test(resultText(remove))) return;
+    if (!GATEWAY_REMOVE_UNSUPPORTED.test(resultText(remove))) {
+      assertExitZero(remove, `cleanup gateway registration ${gatewayName}`);
+    }
+
+    // Remove this fallback once the supported OpenShell floor no longer
+    // includes builds whose local-registration verb was `gateway destroy`.
+    const destroy = await this.command("openshell", ["gateway", "destroy", "-g", gatewayName], {
+      ...options,
+      artifactName: `${artifactName}-legacy-destroy`,
+    });
+    if (destroy.exitCode === 0 || GATEWAY_ALREADY_ABSENT.test(resultText(destroy))) return;
+    assertExitZero(destroy, `cleanup gateway registration ${gatewayName}`);
+  }
+
   async bestEffortCleanupSandbox(
     sandboxName: string,
     options: ShellProbeRunOptions = {},