I searched around yesterday and finally selected 4 GitHub open source skills to create beautiful images for the article. It feels pretty good. Four skills can correspond to different scenarios. Just to clarify, they are not separate skill projects, they are within the skills of Baoyu and Huashu. Another experience is that the drawing effect of Codex directly is worse than calling GPT-Image2 for drawing. Recommend using gpt-image2 1. baoyu-image-cards -Skill Name: ` baoyu image cards'` -Suitable scenarios: social media image cards, knowledge cards, carousel images, and tweet captions. -Usage suggestion: It is suitable to split a tweet into 1-3 key cards, highlighting the title, judgment points, and conclusion. 2. baoyu-cover-image -Skill Name: ` baoyu cover image '` -Suitable scenarios: article cover image, main visual of tweet, title large font image. -Usage suggestion: Suitable for using only one cover image per tweet, focus on enlarging the title and core concepts. 3. huashu-wechat-image -Skill name: huashu wehat image` -Suitable scenes: official account pictures, article illustrations, 16:9 Chinese information pictures, text explanation pictures. -Usage suggestion: Suitable for creating Chinese accurate and structurally clear infographics, such as comparison charts, flowcharts, and field relationship diagrams. 4. huashu-xhs-image -Skill Name: ` huashu xhs image '` -Suitable scenarios: Xiaohongshu style illustrations, vertical social media cards, and large character images of mobile information flow. -Usage suggestion: Suitable for vertical layout, with large titles and few key points, suitable for mobile reading.