TheAlejo160/orchestra ? reverse-engineered prompt

Build me a Claude Code plugin that acts like a smart team of AI agents and sends each part of a task to the model that fits it best. I want it to work with Claude first, but also use other agents I already have installed, like Codex, Gemini, OpenCode, GitHub Copilot, or Kimi, when they are available. It should show a live pane inside Claude Code so I can see what each agent is doing, what files they are touching, and who is waiting on who.

Please also include a simple efficiency mode that keeps answers short, avoids waste, and prefers the smallest complete change. I want a skill file that can also be used on its own in other agent tools, plus a setup script that installs it where possible. Add a master prompt that can build or upgrade the plugin on my machine, and make sure there are tests for the main plugin behavior. If you need to check current docs for the CLIs, go ahead and look them up.