multica-ai/multica ? reverse-engineered prompt

Build me a platform for managing AI coding agents, almost like a Jira but for AI teammates. I want a web app with a board view where I can create coding tasks and assign them directly to an AI agent.

Once an agent is assigned a task, it should work on it autonomously, from start to finish. I don't want to have to babysit the process. The platform should give me real-time updates, maybe the agent can post comments or update its status on the task card so I know what it's doing. It should feel like managing a real developer.

A key feature is that these agents should be able to learn. When one solves a problem, that solution should become a reusable skill for the whole team of agents. The system should run on my own hardware via a local daemon. For the stack, a Next.js frontend and a Go backend would be great, and make it self-hostable with Docker. It should support popular models like Claude.