whoami
Kevin Kulik
Most days I work somewhere between the terminal, code, and infrastructure, tracing why a system behaves differently than expected. Then I document it, repair it, or keep building.
The tools keep changing. What stays is the work underneath: finding out what a system actually does, then changing it so it can still be checked afterwards.
I follow actual behaviour through code, configuration, and data flows before naming a cause.
I make dependencies, ownership, and the unknown parts of a system visible.
I measure the real system instead of relying on screenshots, assumptions, or vendor claims.
I build websites, integrations, and small systems once the problem and its boundaries are understood.
I turn repeatable work into scripts or agent workflows after the manual path has proved reliable.
I run small, understandable systems and document where their boundaries are.
Three things from different years. Where I cannot link anything public, I say so.
In 2014 I took RAT malware apart and traced its command-and-control infrastructure. Two years later I noticed a PIN limit at a local bank that was too short. I worked out what that meant in practice and sent concrete remediation notes.
I reported only what I could prove, and left it at that.
EvidenceThe history is documented in local archive sources. There is no verified public link yet.
I wanted a CSS system that could serve static Astro sites and WordPress with Bricks, with no build step and nothing loaded at runtime. So: tokens, base rules, components, and separate entry points, checked against real projects and a Bricks test environment.
It stays portable and CSS-only. Platform rules live apart, and whatever a project needs differently overrides the vendored core instead of changing it. This site runs on it.
EvidenceVersion 0.2.0 sits in this repository as a verified snapshot. The source repository is not publicly linked at the moment.
I work with several agent runtimes. They all need the same rules, skills, and workflows, and I did not want to build yet another assistant app for that. So it all lives in one portable repository of Markdown, configuration, skills, playbooks, and sync scripts. The vault stays the canonical source for anything that has to last.
That keeps the runtime replaceable. Agents handle the volume; decisions, verification, and responsibility stay with me.
EvidenceStructure, sync paths, and inventory are versioned and documented locally. A public repository link is not planned.
Updated
My homelab is a practice ground and the real thing at once. Whatever runs here, I run myself, and when it breaks I fix it myself.