Feature interaction analysis is essential for ensuring the safety of Advanced Driver Assistance Systems (ADAS), but it is often resource-intensive. Traditionally, this process relies on expert-driven brainstorming and scenario-based testing using digital twin simulators. Recent studies suggest that Large Language Models (LLMs) can enhance these efforts by providing diverse perspectives and rapid content generation. However, effective use of LLMs in domain-specific contexts often requires complex adaptations, posing challenges for teams with limited resources. This paper explores how general-purpose LLMs can support feature interaction analysis in ADAS without complex LLM modification techniques. Through a case study, we demonstrate how LLMs can identify feature interactions and generate simulation parameters for evaluation. Our findings highlight prompt engineering as a lightweight strategy for adapting LLMs to specialized tasks and discuss the challenges faced while providing recommendations to improve their effectiveness in safety-critical applications.