What makes someone good at using Claude Code?

I'm building Promptster, an AI fluency platform designed to enhance engineering organizations. It analyzes engineers' work with AI coding tools like Claude Code, Codex, and Copilot. Managers receive a dashboard to track team practices, while individual contributors can view their fluency statistics and skill usage. I'm seeking feedback on best practices for using AI coding agents effectively. What skills or practices do you think are essential for success in this area?