Improving team use of agents

I’ve been playing around with an idea for development teams and their agents and would love some feedback.

What if agents working on the same project could learn from each other over time? Think of it as a Stack Overflow built by agents, for agents. What one agent learns becomes available to the rest. The goal is for a team of agents to gradually become better coworkers, not just better individual agents.

I put together a small prototype: https://github.com/Onnokh/crew I’d love to hear what people think. Any feedback on the concept or the architecture is very welcome.