fix: report failed CLI provider probes with nonzero status

sandboxes/cli.py

@@ -441,6 +441,7 @@ async def _test():
         providers_to_test = [provider] if provider else ["e2b", "modal", "daytona"]
 
         results = []
+        exit_code = 0
         for p_name in providers_to_test:
             click.echo(f"\n🔬 Testing {p_name}...")
 
@@ -461,19 +462,23 @@ async def _test():
                 # Test execute
                 click.echo("  Executing command...")
                 result = await p.execute_command(sandbox.id, "echo 'Hello from CLI test'")
-                if result.success and "Hello from CLI test" in result.stdout:
+                working = result.success and "Hello from CLI test" in result.stdout
+                if working:
                     click.echo("  ✅ Execution successful")
                 else:
                     click.echo("  ❌ Execution failed", err=True)
+                    exit_code = 1
 
                 # Test destroy
                 click.echo("  Destroying sandbox...")
-                await p.destroy_sandbox(sandbox.id)
+                if not await p.destroy_sandbox(sandbox.id):
+                    raise RuntimeError("Failed to destroy probe sandbox")
                 click.echo("  ✅ Destroyed")
 
-                results.append((p_name, "✅ Working"))
+                results.append((p_name, "✅ Working" if working else "❌ Failed: command test"))
 
             except Exception as e:
+                exit_code = 1
                 click.echo(f"  ❌ Error: {e}", err=True)
                 results.append((p_name, f"❌ Failed: {str(e)[:50]}"))
 
@@ -484,7 +489,9 @@ async def _test():
         for p_name, status in results:
             click.echo(f"{p_name:10} {status}")
 
-    asyncio.run(_test())
+        return exit_code
+
+    sys.exit(asyncio.run(_test()))
 
 
 @cli.command()

tests/test_cli_test_status.py

@@ -0,0 +1,48 @@
+import pytest
+from click.testing import CliRunner
+
+from sandboxes.base import ExecutionResult, Sandbox, SandboxState
+from sandboxes.cli import cli
+
+
+class ProbeProvider:
+    def __init__(self, exit_code, stdout, destroyed=True):
+        self.result = ExecutionResult(exit_code, stdout, "")
+        self.destroyed = destroyed
+
+    async def create_sandbox(self, config):
+        return Sandbox(id="probe", provider="probe", state=SandboxState.RUNNING)
+
+    async def execute_command(self, *args):
+        return self.result
+
+    async def destroy_sandbox(self, sandbox_id):
+        return self.destroyed
+
+
+@pytest.mark.parametrize("exit_code,stdout", [(1, ""), (0, "wrong output")])
+def test_failed_provider_probe_is_reported_as_failure(monkeypatch, exit_code, stdout):
+    monkeypatch.setattr("sandboxes.cli.get_provider", lambda name: ProbeProvider(exit_code, stdout))
+    result = CliRunner().invoke(cli, ["test", "--provider", "e2b"])
+    assert result.exit_code == 1
+    assert "Working" not in result.stdout
+
+
+def test_successful_provider_probe_returns_zero(monkeypatch):
+    monkeypatch.setattr(
+        "sandboxes.cli.get_provider", lambda name: ProbeProvider(0, "Hello from CLI test")
+    )
+    result = CliRunner().invoke(cli, ["test", "--provider", "e2b"])
+    assert result.exit_code == 0
+    assert "Working" in result.stdout
+
+
+def test_failed_probe_cleanup_returns_nonzero(monkeypatch):
+    monkeypatch.setattr(
+        "sandboxes.cli.get_provider",
+        lambda name: ProbeProvider(0, "Hello from CLI test", destroyed=False),
+    )
+    result = CliRunner().invoke(cli, ["test", "--provider", "e2b"])
+    assert result.exit_code == 1
+    assert "Destroyed" not in result.stdout
+    assert "Working" not in result.stdout