With the rapid advancement of humanoid robot technology, precise evaluation of its comprehensive performance is crucial for ensuring stable operation in applications. Here, a multimodal humanoid robot performance evaluation system based on a single-channel tactile-sliding triboelectric nanogenerator (SCTS-TENG) is proposed. SCTS-TENG enables 1D tactile position recognition using the skewness (SK) of the single-channel signal, pressure perception mimicking human skin during sliding, and direction identification. Inspired by the process of ancient silk trade, a performance testing system for humanoid robots is constructed using the SCTS-TENG and a camera. Testing specifications for rotational contact of the dexterous hand and contact-sliding-separation of the robotic arm are defined, and an in-depth analysis is conducted on output signal characteristics under different influencing variables. The SCTS-TENG signals are converted into 2D images using the gramian angular difference field (GADF) method, achieving 97.22% accuracy when used as a standalone modality. This is further fused with camera images, which achieved 75% accuracy as a standalone modality, to develop a multimodal deep learning (DL) system based on the Visual Geometry Group (VGG)19 model. This fusion improves the recognition accuracy to 99.03%. This system provides a low-cost, self-powered, and high-precision solution for sensing and evaluation of humanoid robots.