n8n-io/n8n · error

Sandbox does not support command execution

Error message

Sandbox does not support command execution

What it means

The execute-command tool checks `sandbox.executeCommand` (truthiness) before invoking it. Some sandbox implementations only provide `processes.spawn` and don't implement executeCommand; the tool requires the higher-level executeCommand API, so it refuses to run on such sandboxes rather than calling a missing method.

Source

Thrown at packages/@n8n/agents/src/workspace/tools/execute-command.ts:28

		.input(
			z.object({
				command: z.string().describe('The shell command to execute'),
				cwd: z.string().optional().describe('Working directory for the command'),
				timeout: z.number().optional().describe('Timeout in milliseconds'),
			}),
		)
		.output(
			z.object({
				success: z.boolean(),
				exitCode: z.number(),
				stdout: z.string(),
				stderr: z.string(),
				executionTimeMs: z.number(),
			}),
		)
		.handler(async (input, ctx) => {
			if (!sandbox.executeCommand) {
				throw new Error('Sandbox does not support command execution');
			}
			const env = sandbox.getDefaultCommandEnv?.();
			const result = await sandbox.executeCommand(input.command, undefined, {
				cwd: input.cwd,
				...(env ? { env } : {}),
				timeout: input.timeout,
				abortSignal: ctx.abortSignal,
			});
			return {
				success: result.success,
				exitCode: result.exitCode,
				stdout: result.stdout,
				stderr: result.stderr,
				executionTimeMs: result.executionTimeMs,
			};
		})
		.build();
}

View on GitHub (pinned to 5ac6606e81)

Solutions

1. Use a sandbox that implements executeCommand (e.g. DaytonaSandbox), or don't register the execute-command tool.
2. If building a custom sandbox, implement executeCommand(command, args, options) returning {success, exitCode, stdout, stderr}.
3. Route command execution through runInSandbox() which falls back to processes.spawn when executeCommand is absent.

Example fix

// before: custom sandbox only implements spawn
class MySandbox implements WorkspaceSandbox {
  processes = new ProcManager();
  // no executeCommand
}

// after
class MySandbox implements WorkspaceSandbox {
  processes = new ProcManager();
  async executeCommand(cmd, _args, opts) {
    return /* delegate to processes.spawn */;
  }
}

function supportsExec(sandbox: unknown): sandbox is { executeCommand: NonNullable<unknown> } {
  return typeof sandbox === 'object' && sandbox !== null && 'executeCommand' in sandbox && typeof (sandbox as { executeCommand?: unknown }).executeCommand === 'function';
}
if (!supportsExec(sandbox)) throw new Error('this sandbox cannot run execute-command');

function supportsExec(sandbox: unknown): sandbox is { executeCommand: (...args: unknown[]) => unknown } {
  return Boolean(sandbox) && typeof (sandbox as { executeCommand?: unknown }).executeCommand === 'function';
}

When it happens

Trigger: Registering the execute-command tool against a sandbox whose executeCommand is undefined (a sandbox that only implements processes.spawn). Mixing a spawn-only sandbox with the execute-command tool.

Common situations: A custom WorkspaceSandbox implementation that omits executeCommand; using a minimal/local sandbox shim that only implements spawn; version skew between the tool registry and a sandbox adapter.

Related errors

• Sandbox "${this.name}" has been destroyed
• Sandbox "${this.name}" failed to start (status: ${this.statu
• Daytona sandbox "${this.id}" is not running
• Workspace has no sandbox
• File "${input.path}" is not loaded.

AI-assisted analysis of n8n-io/n8n@5ac6606e81 (2026-08-12). Data as JSON: /api/errors/ba47c893573c076d. Report an issue: GitHub.