English · 简体中文
Skills management, distribution, and rollout — making AI agents pleasant to work with.
I mostly write TypeScript and Python, with some JavaScript, C++, and shell along the way.
I work on the tooling layer around agents. An agent is only as capable as the tools and context it is given, so I spend my time on questions like "how do we make skills easy to use, easy to govern, and easy to distribute" — rather than building yet another chat box.
- The full skills lifecycle: upload → manage → distribute
- MCP integration and the engineering side of agent skills
- Small tools that turn content or video into structured information
| Project | What it is | Stack |
|---|---|---|
| skilldrive | A private skills management SaaS for AI agents: upload → manage → distribute, with multi-tenant isolation, a web console, and a REST API | TypeScript Apache-2.0 |
| awesome-skills | A collection of agent skills plus a Vibe Coding methodology: project kickoff, architecture visualization, content writing | Python |
| StudyMind | A desktop app for classroom learning: local audio/video transcription (SenseVoice ONNX) with AI summaries and mind-map outlines | TypeScript Tauri Python |
| FrameQ | Video → transcript → questions. Turns raw footage into practice questions | TypeScript Apache-2.0 |
| InsightFlow | Turns any web article into AI-crafted questions, so you read deeper | Python MIT |
- chromitrace — Scrapes Q&A pairs and citation links from AI platforms ·
TypeScript - TaskPulse — A master-worker distributed task framework: live monitoring, self-healing, visual control ·
Python - FinScraper — Financial data scraping ·
Python
Open an issue if you want to talk agent tooling.



