Musculoskeletal disorders (MSDs) are among the most common occupational health problems and often develop gradually due to prolonged poor posture and improper movement patterns. Conventional MSD assessment methods typically rely on clinical visits, specialized equipment, or wearable sensors, which limits their accessibility and scalability. This paper presents a low-cost, real-time system for the detection and classification of musculoskeletal disorder risks using video data captured from standard consumer-grade webcams. The proposed framework employs MediaPipe-based human pose estimation to extract skeletal keypoints from video frames, followed by the computation of joint angles and posture-related features. Temporal Convolutional Networks (TCNs) are then utilized to model long-term temporal dependencies in posture dynamics and classify ergonomic risk levels. Experimental evaluation conducted on real-time video streams demonstrates that the proposed TCN-based approach achieves an accuracy of 92.5% while maintaining real-time performance at frame rates between 25 and 30 frames per second. Comparative analysis with Long Short-Term Memory (LSTM) and Graph Convolutional Network (GCN) models further confirms the superiority of TCNs in terms of accuracy, computational efficiency, and real-time suitability. The results indicate that the proposed vision-based system provides an effective, non-invasive, and scalable solution for early MSD risk screening in home, office, and remote work environments.