I build TypeScript products with AI coding agents, and I own the engineering around them: specs, reviews, tests, CI and releases.
I studied network engineering, hold a PMP, and spent seven years in electronics manufacturing (Luxshare, Wingtech, Goertek), the last year leading production-line launches in Vietnam. That is where I learned to run work as a process: fix the cause, not the defect, and stop the line when something is wrong.
What I work with
TypeScript · Node.js · Astro · Tailwind CSS · GitHub Actions · Cloudflare Workers · Python (scripts) · Claude Code and other coding agents
Selected work
- yokoten · open source, MIT, on npm A TypeScript CLI that turns on a coding agent's sandbox and stops the mistakes you can't undo: a wiped folder, a force push, a leaked key. Install it once and it runs in the background; it guards my own agent setup too. Zero runtime dependencies. Runs on Linux, Windows and macOS. Built under its own rules: every change is an issue, a pull request, CI, then a review by a different AI model. 520+ merged pull requests, 2,100+ tests, 63 written decision records, all public.
- cool-workflow · on npm · maintenance mode Turns an AI agent's chat answer into a saved report, with every claim tied to a line of code.
How I work
Issue → pull request → checks → independent review → merge → release. I write down why a decision was made, so the next person (or agent) doesn't have to guess.