With the continuous acceleration of the intelligentization and electrification of automobiles, traditional Hardware-in-the-Loop (HIL) testing is facing problems such as high dependence on human resources, low testing efficiency, and difficulty in knowledge reuse. This paper proposes an intelligent agent architecture that integrates Large Language Model (LLM) and Retrieval-Augmented Generation (RAG). Through hierarchical task decomposition, RAG knowledge enhancement, and code collaborative generation, it achieves efficient generation of test cases and automated test scripts. Experimental results show that this solution can significantly reduce human intervention, lower token consumption, and effectively improve testing development efficiency.