Security protocols play a crucial role in ensuring the confidentiality and integrity of communications over untrusted networks, and their correctness is essential to the overall security of systems. However, the traditional modeling and verification process of security protocols often relies heavily on domain experts, making it time-consuming and difficult to scale. Recent advances in large language models (LLMs) have demonstrated strong capabilities in code generation and semantic reasoning, creating new opportunities for automating the formal modeling of security protocols. However, existing methods still lack domain-specific knowledge and struggle to capture implicit semantics effectively. In this early-stage work, we explore a retrieval-augmented generation (RAG)-based, LLM-assisted modeling framework that automates the generation of formal Tamarin models from natural-language protocol descriptions. The framework first parses protocol text to extract semi-structured key information, then leverages a domain-specific knowledge graph to enhance semantic understanding, and finally integrates a verification-guided feedback loop to improve model reliability and interpretability through verification-guided refinement. Preliminary experiments show a 23.68% improvement in syntactic correctness compared with the state-of-the-art baseline. These results suggest a promising direction for automated and explainable formal modeling of security protocols.