多 Agent 混合层级隔离架构 1.0

OpenClaw 多 Agent 协作架构 - 混合层级 + 物理隔离 + 逻辑隔离的完美组合

This package is documentation + templates for separating a 'writer' and 'media' agent and a local config-check script — overall coherent. Before installing/applying templates: 1) Review the SOUL.md templates (writer/media) to ensure you are comfortable with the enforced constraints and wording (they change agent behavior). 2) Back up any existing SOUL.md and openclaw.json before overwriting or copying files. 3) The config-check.sh reads local OpenClaw files (~/Documents/openclaw/agents, ~/.openclaw/openclaw.json) and checks for the openclaw binary and logs; run it locally to validate environment but inspect its output and exit codes. 4) The recommended softlink approach shares skills between agents — verify that writer does not get unintended access to baoyu-* skills via your filesystem permissions or link locations. 5) No sensitive credentials are requested by this skill, but treat any template that changes agent behavior as granting behavioral control: only apply templates you trust. If you want lower risk, copy templates to a separate test environment first and validate behavior before applying in production.

Type: OpenClaw Skill Name: multi-agent-hybrid-architecture Version: 1.0.0 The skill bundle provides a comprehensive framework for multi-agent orchestration within OpenClaw, defining roles for a coordinator ('Momo') and a creative executor ('Xiaomei'). It includes detailed documentation, personality templates (SOUL.md), and a diagnostic shell script (templates/config-check.sh) designed to verify the local environment setup. While the bundle includes instructions for the agent to perform basic filesystem checks (grep) and provides a shell script for the user to run, these actions are transparently documented and strictly aligned with the stated purpose of managing a multi-agent architecture. No evidence of malicious intent, data exfiltration, or unauthorized execution was found.