asiflow/claude-nexus-hyper-agent-team ? reverse-engineered prompt

Build me a production ready multi agent team for Claude Code that acts like an operating layer on top of the normal coding flow.

I want 32 specialized agents, runtime hooks that enforce the rules, a simple protocol for privileged actions, and a trust system that learns which agents give good advice over time. Include the memory and signal sharing pieces, the review and challenge flow when agents disagree, and the optional shadow mind style helper that can be turned off without breaking the main team.

Please make it feel usable in a real project, with clear setup steps, a root Claude config file, docs, and tests that prove the agents and hooks work together. If you need to look up current Claude Code docs online, go ahead. Also make sure there’s a way to add new specialist agents later when the team finds a gap in coverage.