Luka Živković

At n8n, I work on evaluation systems for AI agents. I care about what happens after a demo looks convincing: how a team can inspect the work, compare results, and decide what to improve.

I also build independent tools and reusable agent skills for inspecting AI runs, developing evaluators, making release decisions, and reviewing skills before installation. The guides here explain how each project works and how to try it.

I like tools with clear responsibilities. A trace system should preserve evidence without judging it. An evaluator should report what it measured without deciding product policy, and a release gate should distinguish a real failure from evidence it couldn’t establish.

The projects published here are independent work and are not n8n products unless a page explicitly says otherwise.