<p>High-performance structures with specific aesthetic features are appealing in various fields and can be tailored through the neural network-assisted topology optimization (NNTO) method. However, the statistical and content-based characteristics of the stylistic features generally lead to overly complex and impractical designs. Therefore, a skeleton-reinforced NNTO method was proposed to design novel structures simultaneously with desired stylistic features and excellent manufacturability. Using the density-based topology optimization method, the convolutional neural network was adopted to introduce stylistic feature constraints. Then the minimum length scale was strictly constrained according to the structural skeleton, which was accurately extracted by the graphic thinning algorithm. The influence of key parameters in the algorithm was deeply studied through a series of numerical examples, and the approximate ranges were determined and verified. Finally, the effectiveness and robustness for designing the stiffest structures with stylistic features and desired minimum length scale were completely proved.</p>