jpjump11/ruflo — reverse-engineered prompt

Build me a project called Ruflo, a multi agent helper for Claude Code and Codex that makes a bunch of AI agents work together instead of acting alone.

I want a simple getting started flow where I can run one command to set it up, then keep using Claude Code normally while Ruflo handles task routing, swarm coordination, memory, and learning from past work in the background. It should support a full install for the complete experience, plus a lighter plugin style mode that just adds slash commands and agent definitions for people who want to try it without changing much.

Please include shared memory across sessions, RAG style knowledge retrieval, reusable workflows, background automation, and secure communication between agents on different machines when enabled. It should feel like a practical agent harness, not just a demo, with good defaults and clear docs so someone can actually use it right away.

If anything is unclear, check the current docs online and keep the behavior aligned with the README.