Artificial-intelligence agents — touted as AI's next wave — could be vulnerable to malicious code hidden in innocent-looking ...