Building an iterative AI skill that learns across MCPs

It changed multiple times but at this moment I have a skill that accumulates knowledge that combines data from different MCPs. As an example, I start working on CVR for a specific country, reviewing keywords/description then set a reminder to check the improvements after ~4 weeks to check it again - MCP simplifies that process. I think the most interesting that I found that after each session I ask Claude to update the skill with what he learned and it constantly improves that skill with more and more knowledge. Works amazingly well. That skill is located in the project repo. Later, I’ll refine it and create a shared skill for the projects. It’s an iterative process.