The Lab When Agent Skills Evolve: The Need for a Layer of Control What we learnt from building Rhizomatic, a human-in-the-loop proof of concept for reviewing, versioning and rolling back changes to reusable agent skills. AgenticInfrastructure