fix: error handling — no more silent failures

cmd/creds.go

@@ -91,7 +91,9 @@ func createGWSSecret(ctx context.Context, kc *k8s.Client) error {
 		fmt.Println("  GWS: export failed (skipping)")
 		return nil
 	}
-	os.WriteFile(credFile, out, 0o600)
+	if err := os.WriteFile(credFile, out, 0o600); err != nil {
+		return fmt.Errorf("writing gws credentials: %w", err)
+	}
 
 	args := []string{"create", "secret", "generic", "openshell-gws",
 		"--from-file=credentials.json=" + credFile}

cmd/deploy.go

@@ -74,20 +74,24 @@ func deployRemote(harnessDir string, gw gateway.Gateway, kubeconfig string) erro
 	kc := k8s.New(kubeconfig, namespace)
 	kcNoNS := k8s.New(kubeconfig, "")
 
-	// Step 1: Namespace
+	// Step 1: Namespace (idempotent — ignore AlreadyExists)
 	fmt.Println("=== Step 1: Creating namespace ===")
 	kcNoNS.RunKubectl(ctx, "create", "ns", namespace)
-	kcNoNS.RunKubectl(ctx, "label", "ns", namespace,
+	if _, err := kcNoNS.RunKubectl(ctx, "label", "ns", namespace,
 		"pod-security.kubernetes.io/enforce=privileged",
 		"pod-security.kubernetes.io/warn=privileged",
-		"--overwrite")
+		"--overwrite"); err != nil {
+		return fmt.Errorf("labeling namespace: %w", err)
+	}
 
 	// Step 2: Sandbox CRD
 	fmt.Println("=== Step 2: Installing Sandbox CRD ===")
-	kcNoNS.RunKubectlPassthrough(ctx, "apply", "-f",
-		"https://github.com/kubernetes-sigs/agent-sandbox/releases/latest/download/manifest.yaml")
+	if err := kcNoNS.RunKubectlPassthrough(ctx, "apply", "-f",
+		"https://github.com/kubernetes-sigs/agent-sandbox/releases/latest/download/manifest.yaml"); err != nil {
+		return fmt.Errorf("installing sandbox CRD: %w", err)
+	}
 
-	// Step 3: OpenShift SCCs
+	// Step 3: OpenShift SCCs (best-effort — oc may not exist on non-OpenShift)
 	fmt.Println("=== Step 3: Granting OpenShift SCCs ===")
 	for _, sa := range []string{"openshell", "openshell-sandbox", "default"} {
 		kc.RunOC(ctx, "adm", "policy", "add-scc-to-user", "privileged", "-z", sa, "-n", namespace)
@@ -98,7 +102,7 @@ func deployRemote(harnessDir string, gw gateway.Gateway, kubeconfig string) erro
 		"--serviceaccount=agent-sandbox-system:agent-sandbox-controller")
 
 	// RBAC for launcher
-	kc.ApplyYAML(ctx,
+	if err := kc.ApplyYAML(ctx,
 		map[string]any{
 			"apiVersion": "v1",
 			"kind":       "ServiceAccount",
@@ -129,7 +133,9 @@ func deployRemote(harnessDir string, gw gateway.Gateway, kubeconfig string) erro
 				"apiGroup": "rbac.authorization.k8s.io",
 			},
 		},
-	)
+	); err != nil {
+		return fmt.Errorf("applying launcher RBAC: %w", err)
+	}
 
 	// Step 4: Helm install
 	fmt.Println("=== Step 4: Deploying gateway via Helm ===")
@@ -140,7 +146,7 @@ func deployRemote(harnessDir string, gw gateway.Gateway, kubeconfig string) erro
 
 	appsDomain, err := kcNoNS.GetJSONPath(ctx, "ingresses.config.openshift.io/cluster", "{.spec.domain}")
 	if err != nil || appsDomain == "" {
-		return fmt.Errorf("could not determine OpenShift apps domain")
+		return fmt.Errorf("could not determine OpenShift apps domain — is this an OpenShift cluster? (kubectl get ingresses.config.openshift.io cluster)")
 	}
 	routeHost := fmt.Sprintf("gateway-openshell.%s", appsDomain)
 
@@ -157,11 +163,15 @@ func deployRemote(harnessDir string, gw gateway.Gateway, kubeconfig string) erro
 	if sps := os.Getenv("SANDBOX_PULL_SECRET"); sps != "" {
 		helmArgs = append(helmArgs, "--set", "server.sandboxImagePullSecrets[0].name="+sps)
 	}
-	kc.RunHelm(ctx, helmArgs...)
+	if _, err := kc.RunHelm(ctx, helmArgs...); err != nil {
+		return fmt.Errorf("helm install failed: %w", err)
+	}
 
 	// Wait for gateway
 	fmt.Println("=== Waiting for gateway ===")
-	kc.RunKubectlPassthrough(ctx, "rollout", "status", "statefulset/openshell", "--timeout=300s")
+	if err := kc.RunKubectlPassthrough(ctx, "rollout", "status", "statefulset/openshell", "--timeout=300s"); err != nil {
+		return fmt.Errorf("gateway rollout failed: %w", err)
+	}
 
 	// Step 5: Route
 	fmt.Println("=== Step 5: Creating OpenShift route ===")
@@ -195,7 +205,9 @@ func deployRemote(harnessDir string, gw gateway.Gateway, kubeconfig string) erro
 		}
 	}
 
-	gw.GatewayAdd(gatewayURL, gatewayName, true)
+	if err := gw.GatewayAdd(gatewayURL, gatewayName, true); err != nil {
+		return fmt.Errorf("registering gateway %s: %w", gatewayName, err)
+	}
 
 	// Extract mTLS certs
 	home, _ := os.UserHomeDir()

cmd/new.go

@@ -123,10 +123,12 @@ func newRemote(harnessDir string, gw gateway.Gateway, profileName, sandboxName s
 	if err != nil {
 		return fmt.Errorf("creating config configmap: %w", err)
 	}
-	kc.RunKubectlOpts(ctx, k8s.KubectlOpts{
+	if _, err := kc.RunKubectlOpts(ctx, k8s.KubectlOpts{
 		Args:  []string{"apply", "-f", "-"},
 		Stdin: strings.NewReader(out),
-	})
+	}); err != nil {
+		return fmt.Errorf("applying config configmap: %w", err)
+	}
 
 	// 2. ConfigMap from env (conditional)
 	envContent := cfg.BuildSandboxEnv()
@@ -135,14 +137,16 @@ func newRemote(harnessDir string, gw gateway.Gateway, profileName, sandboxName s
 			"--from-literal=sandbox.env="+envContent,
 			"--dry-run=client", "-o", "yaml")
 		if err == nil {
-			kc.RunKubectlOpts(ctx, k8s.KubectlOpts{
+			if _, err := kc.RunKubectlOpts(ctx, k8s.KubectlOpts{
 				Args:  []string{"apply", "-f", "-"},
 				Stdin: strings.NewReader(out),
-			})
+			}); err != nil {
+				return fmt.Errorf("applying env configmap: %w", err)
+			}
 		}
 	}
 
-	// 3. Clean up old job
+	// 3. Clean up old job (best-effort — may not exist)
 	jobName := "sandbox-" + cfg.Name
 	kc.RunKubectl(ctx, "delete", "job", jobName, "--force", "--grace-period=0")
 	kc.RunKubectl(ctx, "delete", "pod", "-l", "job-name="+jobName, "--force", "--grace-period=0")
@@ -200,9 +204,11 @@ func newRemote(harnessDir string, gw gateway.Gateway, profileName, sandboxName s
 	logCmd.Stderr = os.Stderr
 	logCmd.Start()
 
-	// 7. Poll job status
-	for {
-		jobStatus, _ := kc.RunKubectl(ctx, "get", "job", jobName,
+	// 7. Poll job status (10 min timeout)
+	var jobStatus string
+	deadline := time.Now().Add(10 * time.Minute)
+	for time.Now().Before(deadline) {
+		jobStatus, _ = kc.RunKubectl(ctx, "get", "job", jobName,
 			"-o", "jsonpath={.status.conditions[0].type}")
 		if jobStatus == "Complete" || jobStatus == "Failed" || jobStatus == "SuccessCriteriaMet" {
 			break
@@ -216,14 +222,14 @@ func newRemote(harnessDir string, gw gateway.Gateway, profileName, sandboxName s
 	}
 
 	fmt.Println()
-	jobStatus, _ := kc.RunKubectl(ctx, "get", "job", jobName,
-		"-o", "jsonpath={.status.conditions[0].type}")
 	if jobStatus == "Complete" || jobStatus == "SuccessCriteriaMet" {
 		fmt.Printf("Sandbox ready. Connect with: harness connect %s\n", cfg.Name)
-	} else {
-		fmt.Println("Launcher failed.")
+		return nil
 	}
-	return nil
+	if jobStatus == "" {
+		return fmt.Errorf("launcher job timed out — check: kubectl logs -n %s -l job-name=%s", namespace, jobName)
+	}
+	return fmt.Errorf("launcher job failed (status: %s) — check: kubectl logs -n %s -l job-name=%s", jobStatus, namespace, jobName)
 }
 
 func newLocal(opts newLocalOpts) error {
@@ -280,7 +286,9 @@ func newLocal(opts newLocalOpts) error {
 
 	// 5. Stage files
 	harnessUploadDir := "/tmp/openshell"
-	os.RemoveAll(harnessUploadDir)
+	if err := os.RemoveAll(harnessUploadDir); err != nil {
+		return fmt.Errorf("cleaning staging dir: %w", err)
+	}
 	if err := profile.StageHarnessDir(cfg, harnessUploadDir); err != nil {
 		return fmt.Errorf("staging files: %w", err)
 	}
@@ -311,11 +319,11 @@ func newLocal(opts newLocalOpts) error {
 			return nil
 		}
 
-		fmt.Printf("  Attempt %d failed (supervisor race), retrying in 5s...\n", attempt)
-		gw.SandboxDelete(cfg.Name)
+		fmt.Printf("  Attempt %d failed: %v, retrying in 5s...\n", attempt, err)
+		gw.SandboxDelete(cfg.Name) // best-effort cleanup
 
 		if attempt == 5 {
-			return fmt.Errorf("failed after 5 attempts")
+			return fmt.Errorf("sandbox create failed after 5 attempts: %w", err)
 		}
 		time.Sleep(opts.retrySleep)
 	}

internal/profile/profile.go

@@ -106,7 +106,9 @@ func stageGWSCreds(harnessDir string) error {
 	}
 	clientSecret := filepath.Join(gwsConfigDir, "client_secret.json")
 	if data, err := os.ReadFile(clientSecret); err == nil {
-		os.WriteFile(filepath.Join(harnessDir, "client_secret.json"), data, 0o600)
+		if err := os.WriteFile(filepath.Join(harnessDir, "client_secret.json"), data, 0o600); err != nil {
+			return fmt.Errorf("writing client_secret.json: %w", err)
+		}
 	}
 
 	fmt.Println("  GWS: exported")

test/test-flow.sh

@@ -204,6 +204,10 @@ test_ocp() {
   echo "=== test-flow: ocp ($mode) [$IMPL] ==="
 
   if $REUSE_GATEWAY; then
+    # Ensure OCP gateway is selected (error scenarios may have switched to local)
+    OCP_GW=$("$CLI" gateway list 2>/dev/null | strip_ansi | awk '/-remote-/ {gsub(/^\*/, ""); print $1; exit}')
+    [[ -n "$OCP_GW" ]] && "$CLI" gateway select "$OCP_GW" 2>/dev/null || true
+
     step "teardown sandboxes+providers" "$HARNESS" teardown --sandboxes --providers
     # Deploy only if gateway is not reachable
     if ! "$CLI" inference get &>/dev/null; then