Senior DevOps/SRE who treats AI as a first-class part of the toolchain — I design the guardrails, skills, and verify loops that let agents propose and build code safely, while a human owns the merge. Below is one thing that proves it: a server-authoritative game platform I built end-to-end with an agent-driven pipeline.
A personal build that exercises the whole skill set at once: Go backend, embedded Lua sandbox, Postgres, Unity client, OTel observability — and an agent toolkit (skills + contracts + a live verify loop) that ships a new game on top of it.
Server-authoritative game platform. Game rules are Lua scripts stored in Postgres and run in a sandboxed VM — ship a new game's logic without a client build. The SDK is a Unity UPM package; the UI is built from JSON contracts; the whole loop is driven and verified by agents.
game.* bridge is exposed.ContractBuilder instantiates it, the verify loop diffs the live tree against it.battle_sessions.state JSONB + idempotency keys) serves every game.spec → contract → assets → build → verify) plus a per-game playtest skill.{ "screen":"WinBanner", "template":"popup.banner",
"theme":"kawaii", "title":"Level Clear!",
"buttons":[{"id":"banner-retry"},{"id":"banner-levels"}] }
// ContractBuilder.Build() → live UI; UiDump asserts the ids.