At term (and not even far term) - LLMs will be able to churn up their own "skills" using their sandbox code environments - and possibly recycle them through con

At term (and not even far term) - LLMs will be able to churn up their own "skills" using their sandbox code environments - and possibly recycle them through context on a per-user basis. While I like the flexibility of deploying your own skills to claude for use org-wide, this really feels like what MCP should be for that use case, or what built-in analysis sandbox should be. We haven't even gone mainstream with MCP and there are already 10 stand-ins doing roughly the same thing with a different twist. I would have honestly preferred they called this embedded MCP instead of 'skills'.