ruvnet/ruflo ? reverse-engineered prompt

Build me an advanced AI agent orchestration platform, let's call it Ruflo. The main idea is to create a system that can manage and coordinate teams of specialized AI agents to work together on complex software development tasks. It should be built with TypeScript and Node.js.

The platform needs to support a whole library of pre-built agents for different roles, like a coder, a security analyst, a code reviewer, and a tester. A key feature is that these agents can work together in "swarms," communicating and collaborating to achieve a goal. The system should be smart enough to route tasks to the best agent for the job.

I also want it to be self-learning. It should observe outcomes, remember successful patterns, and continually optimize its own performance. It needs to connect to various LLM providers, especially Claude, but also GPT and local models via Ollama. The goal is to create an enterprise-grade framework that lets you deploy autonomous AI systems. It should have a clean command-line interface to get started.