The resurgence of Monkeypox as a worldwide public health problem, for which no treatment is available, poses a remarkable diagnostic challenge, especially in mimicking the morphological lesions of other infectious skin diseases, such as chickenpox and Measles. Although Polymerase Chain Reaction (PCR) is the gold standard for diagnosis due to its high sensitivity, it is not widely used in low-resource settings due to cost and infrastructure constraints. To fill this gap, in this research, we propose a Deep Learning (DL) model for the multi-class classification of skin lesions into four classes: Monkeypox, Chickenpox, Measles, and Normal skin. We curated and pre-processed a clinically validated dataset consisting of 2,773 images, which were subjected to extensive data augmentation to improve the generalisation capabilities of the model. We performed the experiments using transfer learning with three popular CNN models: VGG16, ResNet50, and InceptionV3. From comparative experimentation, we identified ResNet50 as the best model, outperforming other tested models with an average cross-validation accuracy of 83.3% and a final test accuracy of 95.2%, having higher precision, recall, and F-1 scores across all classes. In order to transfer this experimental research toward clinical application, a web-based diagnostic tool was developed using the proposed model. This easy-to-use solution provides reliable, cost-effective testing on desktop and mobile platforms, and the results are promising for the prospects of AI-assisted RPCVs in telemedicine at a community-based healthcare screening level