code-yeongyu/oh-my-openagent ? reverse-engineered prompt

I want to build a tool that can orchestrate different AI models to work on coding tasks. The main idea is that I don't want to be locked into just one provider like OpenAI or Anthropic. Instead, this tool should be able to use the best model for the job—maybe Claude for orchestration, GPT for reasoning, and Gemini for creativity, all working together on the same problem.

It should be a command-line tool with a really simple interface. I'm imagining a single powerful command, like ultrawork, where I can give it a big task (like refactoring a whole project) and it will just keep working until it's done. Think of it as an open-source, power-user alternative to something like Cursor. Let's build it with TypeScript. You can look up current docs for connecting to the various model APIs as needed.