Product engineer building agent workflows and reliable software systems.
Turning recurring engineering friction into reusable tools, rules, and ways of working.
Kuala Lumpur · TypeScript · AI agents · Developer experience
Writing ·
Projects
I work where product engineering, AI agents, and developer experience meet. I care less about making agents look busy and more about helping them make better decisions: clarify intent, respect system boundaries, preserve project knowledge, and learn safely from human feedback.
Right now, most of that thinking lives in Codeartz Skills, a Chinese-first collection of focused Agent Skills extracted from real engineering collaboration problems.
| Project | Why it exists |
|---|---|
| Codeartz Skills | Reusable Agent Skills for unclear intent, mixed requirements, missing project knowledge, unreliable instruction manuals, and feedback that should improve future decisions. |
| Codeartz Illustrations | A Codeartz AI Cat illustration skill for turning Chinese articles, workflows, and ideas into clean hand-drawn visual explanations. |
| Green Energy City | An interactive 3D energy-operations cockpit built with React, TypeScript, and Three.js, with explicit layout contracts and automated spatial verification. |
- Make the real intent discussable before turning it into a polished specification.
- Read the code before claiming how a system behaves.
- Keep authorized changes separate from unknown, conflicting, and unchanged behavior.
- Prefer small tools with explicit boundaries over frameworks that try to own the whole workflow.
- Turn useful human corrections into durable project guidance without creating a pile of duplicate rules.
Most days: TypeScript, React, Angular, Vue, Node.js, agent tooling, RAG, browser automation, and interactive 3D web experiences.
I still enjoy breaking things. I just leave better evidence, boundaries, and recovery paths now.