Skip to content
View marko-builds's full-sized avatar

Block or report marko-builds

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
marko-builds/README.md

Marko Stankovic

AI engineer. I build with autonomous coding agents and ship what they build.

markostankovic.org · LinkedIn · contact@markostankovic.org


Claude TypeScript Python React C# Linux


How I work

I run a personal AI operating system I built on Claude Code. It drives my whole build pipeline: 40+ reusable automation skills, multi-agent orchestration, custom subagents, and integrations into email, calendar, and GitHub.

The part I like most runs while I sleep. An unattended overnight loop picks the next unblocked issue off a backlog, implements it test-first, and has to pass an automated code review before anything gets committed. Whole feature branches ship this way.

Driving agents is the easy half. The harder half is judging what they produce, so the discipline around them is the real work: test-driven tracer bullets, a zero-context red-team review on every non-trivial plan, and a rule that nothing counts as done without fresh passing evidence.

It started as Nate Herk's AIS-OS starter kit, three skills and a framework. It is now around fifty, plus hooks, an issue backlog, and the overnight loop above.

Most of that system is private, because it runs my life and not just my code. So what you see below are the pieces I pulled out of it, plus the products it helped build.


Products

DeployLog: release notes from your deploys. Full-stack SaaS built end-to-end. A hosted page, an embeddable widget, an npm CLI that turns git commits into release notes with an LLM, and a GitHub Action. Launching.

Habitagram: habit tracking as a journey. A React Native app where your habits move a boat across illustrated landscapes that shift with the time of day. The art is rendered from code, not drawn. Near launch.


Open source

skill-vibe-test: does your Claude Code skill still steer the model after 10 turns, or has it quietly stopped working? Scripted conversation, LLM judge, decay chart. I wrote it because I kept guessing at the answer. One of my own skills failed it.

duskpaper: generative animated wallpapers for Wayland, rendered on your machine, at your resolution, with your seed. The public face of the procedural engines behind Habitagram's art.

BlenderBridge: one-click validated Blender to Unreal asset ingest. Naming enforcement, normal maps, lightmap UVs, and a pre-flight report. Hybrid C++ and Python editor plugin.


Background

Three years on game systems in Unity and Unreal before this (BSc Game Development, ex-Replai): save, survival, and inventory systems behind clean interfaces, documented with ADRs and tests. The habits came with me. The engine changed.


Toolbox

Claude Code · Anthropic API · MCP · multi-agent workflows · evals

TypeScript / Node (Next.js, Supabase) · Python · React Native · C# · C++ · Unity · Unreal Engine 5 · Blender

CLI-first on Arch Linux, automation everywhere.

Pinned Loading

  1. deploylogdev/deploylog deploylogdev/deploylog Public

  2. deploylogdev/action deploylogdev/action Public

    TypeScript

  3. deploylog-cli deploylog-cli Public

    TypeScript

  4. portfolio portfolio Public

    Personal portfolio website — game developer

    Astro

  5. duskpaper duskpaper Public

    Generative animated wallpapers for Wayland/Hyprland, rendered on your machine, at your resolution, with your seed

    Python

  6. skill-vibe-test skill-vibe-test Public

    Does your Claude Code skill still steer the model after 10 turns? Live degradation probe: scripted conversation, LLM judge, decay chart.

    Python