As large language models (LLMs) reshape software development, introductory computing (CS1) courses must adapt. We present and evaluate an LLM-integrated CS1 course that emphasizes a broader range of computing skills than traditional CS1 courses, including problem decomposition, testing, debugging, and interacting with LLMs and prompting. We investigate (1) what skills students use when coding with LLMs and (2) how pedagogy impacts these skills. Our methods include observation studies, surveys, and course interventions.