Great reminder that giving AI more access does not always mean better results. As AI tools become more integrated into development workflows, security practices like least privilege become even more important. Just like we carefully define permissions for human team members, we should also think about what data, systems, and actions an AI agent truly needs to perform its task. A thoughtful balance between capability and control will be key for building reliable AI-powered workflows. Simple principle, but a very important one for anyone working with AI systems today. Thanks for highlighting this!