huangrichao2020/pretty-skills

Pretty-skills is a shared library of reusable AI agent skills, meant to work across different AI coding assistants such as MiniMax Code, Claude Code, Codex, Cursor, and Windsurf. Instead of installing a new one-off tool every time you need a capability, the idea is to build up a single growing collection of proven skills that all of your agents can share. The library covers more than 50 skills across around 15 areas, including things like agent design, content writing, visual creation such as slide decks and covers, financial and stock analysis, programming helpers, and industry research. Each skill in the collection is meant to have been used in real practice for several months before being added, and comes with a real example case, plain-language documentation, and output in formats like PDF or PowerPoint that both people and AI can read. When a user wants to learn something new, they can simply tell their agent to learn it, and the agent will pull the source material, extract the key method, file it into the right category folder, and turn it into a new reusable skill, so the library keeps growing on its own from real usage. The project also describes a daily self-improvement process where usage feedback is reviewed and folded back into future versions of the library. A concrete example given in the README shows a user handing an agent a stock ticker and a request for a deep-dive analysis, and the agent uses the library's stock analysis and slide-deck skills to gather data from multiple sources and produce a finished PowerPoint presentation with charts and Chinese-language commentary, with no manual formatting needed afterward. The project is written in Chinese, encourages contribution through forking and pull requests with automated checks and merging, and is released under the MIT license.