Skin cancer, particularly melanoma, is an overseeing basis of death rate worldwide, with early detection being critical for improving patient health. Dermatoscopy, an imaging technique, is widely used for skin lesion analysis; however, its diagnostic accuracy heavily relies on the expertise of dermatologists, leading to variability and subjectivity. To address these challenges, this research proposes MelanoXAI, an explainable AI framework for the automated detection of melanoma and other skin diseases using dermatoscopic images. Refined to the ISIC Archive and HAM10000 datasets, the framework is a pre-trained EfficientNet model that achieves high diagnostic accuracy while preserving computational efficiency. The Gradient-weighted Class Activation Mapping (Grad-CAM) generates visual heatmaps highlighting regions of SHapley Additive exPlanations (SHAP), which is used to provide feature-level explanations for single predictions. The proposed algorithm is verified using clinical annotations, and the alignment of AI explanations with expert insights is evaluated using quantitative measures like the Dice coefficient and Intersection of Union (IoU). The Experimental analysis demonstrates that MelanoXAI outperforms state-of-the-art methods in accuracy and interpretability, achieving a result F1-score of 92.5% and an AUC-ROC of 94.8% on the ISBI 2016 challenge dataset. By combining advanced deep learning with explainable AI techniques, this research bridges the extended gap between XAI technology and clinical practice, providing dermatologists with a reliable, transparent, and efficient tool for skin cancer detection.