Nexus Multi-Agent Collaboration Hub

Problem

Multi-agent systems often fail because role boundaries are unclear, process artifacts are scattered, and cross-agent communication lacks a stable format. Nexus Multi-Agent Collaboration Hub targets OpenClaw workflows and splits task planning, search, code execution, documentation, monitoring, and user notification into six fixed roles: main, gatherer, coder, writer, monitor, and notifier.

How it works

The skill uses a central coordination model, fixed workspace layout, and enforced output format:
- Coordinator Jarvis/main handles task decomposition, global coordination, reviews, and memory management.
- Perceiver Xunyi/gatherer handles search, source collection, and requirement clarification.
- Executor Zhicheng/coder handles code generation, debugging, and technical feedback.
- Writer Shuyan/writer handles documentation, meeting notes, and expression polish.
- Monitor Muguan/monitor handles task monitoring, alerts, and logging.
- Notifier Jianxin/notifier handles result delivery and user communication.

Key steps include startup self-check, routing external messages through the coordinator, wrapping internal communication in code blocks, storing artifacts under directories from /0_config to /6_output, and retaining user preferences, long-term agent memory, task history, and rules under /memory.

Boundaries

This skill fits OpenClaw workflows that need multi-role collaboration, process traceability, and a fixed delivery format. It may be heavier for one-shot prompts, short answers, or tasks without persistent memory. Users should pay attention to role permissions, workspace write access, log volume, and how memory files are actually maintained.