Early and accurate identification of common skin conditions is essential for appropriate therapy and better clinical outcomes. However, diagnostic expertise is often limited in resource-constrained healthcare or remote settings. We present a robust deep learning approach based on a custom convolutional neural network (CNN) to automatically classify five prevalent skin diseases (acne, contact dermatitis, nail fungus, scabies, and urticaria) from clinical images. A curated dataset of 8,900 clinical images (1,780 per class) was compiled, with standardized resizing, normalization, and extensive augmentation applied to simulate real-world variability and enhance generalization. The CNN architecture consists of five sequential convolutional blocks with progressively increasing filters to extract high-level features, accompanied by dropout layers to mitigate overfitting, and includes a fully connected layer prior to the SoftMax output. On evaluation, the model achieved an overall classification accuracy of 99%, demonstrating robust performance and high diagnostic precision across all categories. It notably outperformed conventional machine learning classifiers as well as state-of-the-art pretrained deep networks on the same task. This exceptional performance, combined with strong generalization capabilities, underscores the model’s reliability and suitability for deployment in real-world clinical practice and tele dermatology settings. In such scenarios, rapid and accurate screening for multiple skin conditions can significantly improve patient care and enable earlier interventions.