Skills are fundamentally different. The only similarity I can see is perhaps the fact that in both cases an LLM is used to find the right skillset.

From my reading, someone could use ChatBotKit to build something like Claude Code skills, but the implementer would have to figure out the non-trivial mechanism to auto-invoke the right set of skills for a task, where Claude Code gives you that part out of the box.