Multi Agent Config Manager

多代理编排引擎 - 目标驱动的深度研究与项目协作系统。支持任务分解、分支执行、验证审核、返工迭代、智能决策。遵循第一性原理，实现主代理与分支代理的双向通信。触发词：多代理、multi-agent、代理编排、深度研究、目标分解、任务委派、工作流、agent orchestrate、multi agent

This skill appears to do what it says (a multi‑agent orchestrator) and contains substantial source code rather than being instruction‑only. Before installing or running it: 1) Back up any important files under ~/.openclaw or your workspace; the skill will create/modify files there and its clean/archive commands delete files in shared/. 2) Run multi-agent-orchestrator check_env and review the returned list; prefer dry_run modes where available. 3) Inspect .multi-agent-profiles.json and other generated configs before accepting auto‑configure; consider creating the directory structure yourself and supplying minimal configs first. 4) Be cautious running long tasks: PollMonitor has an identified busy‑loop risk (may spin if not used with the intended agent yield) — test on a safe task and adjust pollInterval/timeout. 5) No external credentials are required by the skill, but do not place secrets inside the workspace that you do not want archived/copied. If you need lower risk, run the skill in an isolated environment or container and review the code paths for archive/clean before giving it access to production data.

Type: OpenClaw Skill Name: multi-agent-config-manager Version: 7.2.0 The skill bundle is a highly complex and well-documented multi-agent orchestration engine designed for task decomposition, parallel execution, and result aggregation within the OpenClaw environment. It utilizes platform-specific APIs like sessions_spawn and subagents to manage workflows. While the system possesses broad file system capabilities—including automated workspace directory creation (lib/config_validator.js), result archiving, and temporary file cleanup (lib/archiver.js)—these actions are strictly aligned with its stated purpose of managing research projects. The code includes robust error handling, quality validation (lib/validator.js), and detailed audit logs (AUDIT_v8.1.md), indicating a focus on operational reliability and transparency. No evidence of data exfiltration, unauthorized persistence, or malicious command execution was detected.