haoaaa-111/taoketong

Taoketong, which the README translates as Skip Class Master, is a Chinese-language tool for university students that takes a class schedule and produces a personal plan for which classes to skip and which ones to attend. The user uploads a screenshot of their timetable, a short profile of themselves, school information they have gathered, and details of the courses they already know about. The system then returns a complete and concrete plan. The stated goal is not to encourage skipping class but to reduce the time students spend agonizing over the decision. The README lays out scenarios such as an early morning class after a bad night of sleep, or a slow afternoon lecture where the student keeps switching between going and not going. The output is a three way classification per class period: can skip, sign in then leave, must attend, or find a substitute. The user makes the final call on the harder cases. Under the hood, the project is a Next.js 15 app with strict TypeScript and Zod validation, and it uses an OpenAI compatible API. The README skips a traditional database and stores everything as structured Markdown tables on disk through a module called fs-store. The agent stack has four parts. Modeler estimates roll call risk per class from teacher type, school specific rules, and past events. Supervisor turns the Modeler output into the full plan using an eight layer prompt. Memory Agent reads back user feedback about what actually happened and updates the course data and teacher templates. School Skill Generator turns scraped school guides from Jisuke or Xiaohongshu into a per school skill file. The project also ships as an Agent Skill at skills/skipclass that can be loaded into Claude Code, OpenCode, or Codex by copying it into the tool's skills folder, after which the user can say something like analyze my class schedule to trigger it. Self hosting needs npm install, an LLM API key, base URL, and model name in env, then npm run dev. The project is MIT licensed.