Skin cancer, a major global health concern, demands early detection for optimal patient outcomes. Traditional methods relying on subjective dermatological examinations often prove time-consuming and susceptible to human error. Fortunately, recent advancements in deep learning, particularly Convolutional Neural Networks (CNNs), have showcased positive outcomes in improving skin cancer classification accuracy. Building upon this success, this research delves into the potential of integrating transformer layers within a Convolutional Neural Network (CNN) framework for further advancements. By explicitly capturing intricate spatial dependencies and enhancing feature extraction capabilities, our proposed hybrid model seeks to surpass the limitations of individual architectures and offer a robust and comprehensive tool for skin disease classification. This study aims to demonstrate the model's efficacy on a large-scale skin cancer dataset, evaluating its performance against established approaches and offering valuable perspectives on the capacity of hybrid architectures for improved skin cancer diagnosis.