ALI-Agent: Assessing LLMs'Alignment with Human Values via Agent-based Evaluation