As global mortality rates rise alongside an increasing incidence of skin cancer, particularly Basal Cell Carcinoma (BCC), the development of effective automated detection strategies gains urgency. Traditional diagnosis of BCC relies heavily on manual inspection of skin lesions, an approach limited by subjectivity, time constraints, and the invasive nature of biopsy procedures. To address these challenges, this study introduces the two-stage cooperative co-evolution (2SCC)-Criptor, a novel two-stage feature manipulation method that employs genetic programming and co-operative co-evolution for automated BCC identification. The first stage generates an ensemble of models that collaboratively extract discriminative features from decomposed colour channels of skin lesion images, while the second stage constructs additional features to enhance classification performance. The efficacy of the method is evaluated using standard machine learning classifiers, demonstrating statistically significant superiority over traditional and contemporary approaches in the field. Further analysis reveals that the model’s success stems from the synergistic integration of colour channels rather than individual channel contributions, with remarkably uniform utilisation of LAB colour space components. The findings underscore 2SCC-Criptor’s potential in enhancing BCC diagnostic processes while maintaining interpretability through evolved genetic programming individuals.