Skynarc publishes articles written out of curiosity — about AI, and the questions it raises that are worth working through by hand.
Much of what is written about AI is either a pitch or a summary of someone else's pitch. The articles here exist for a simpler reason: a question seemed worth answering properly, and the best way to answer it was to build something small, test it, and write down what happened.
There is no product, no service and nothing for sale. Skynarc does not take on projects or clients. The articles are the whole point.
The authors are still learning. The labs are small, the conclusions are provisional, and some choices will turn out to be wrong. That is part of writing in the open.
Each article starts from something worth understanding. How should an AI agent prove who it is? Who is it acting for? What happens when it misbehaves?
The question is tested on real infrastructure — real identity providers, real cloud permissions, real audit logs — not on slides.
What was built, what the results showed, and where the limits are. Screenshots and outputs are included so readers can check the reasoning.