메뉴
Multi-agent orchestration framework for OpenClaw. Define roles, route tasks, manage state, and coordinate agent teams using structured YAML configs and a proven communication protocol. Based on the YiHui team real production setup (小语/小美/大龙).
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | agent-swarm |
| description | Multi-agent orchestration framework for OpenClaw. Define roles, route tasks, manage state, and coordinate agent teams using structured YAML configs and a proven communication protocol. Based on the YiHui team real production setup (小语/小美/大龙). |
| triggers | ["多 agent 协作","编排","任务路由","agent 团队","角色定义","task routing","agent coordination","multi-agent orchestration"] |
| tags | ["multi-agent","orchestration","workflow","coordination","openclaw"] |
| compatibility | openclaw |
| license | MIT |
| source | https://github.com/1yihui/YiHui |
Multi-agent orchestration framework for OpenClaw. Define roles, route tasks, manage state, and coordinate agent teams using a proven communication protocol.
YiHui 团队经验：最好的编排不是"一个万能 Agent"，而是专角色 + 清晰协议 + 状态可见。
三个原则：
User Input
↓
[Router Agent] ← 解析意图，判断类型
↓
[Specialist Agent(s)] ← 执行子任务
↓
[Critic Agent] ← 审查结果，决策是否通过
↓
[Master Agent] ← 最终输出，兜底
职责：调度、审查、路由、兜底
特征：
行为规则：
职责：特定领域的深度执行
特征：
行为规则：
职责：审查 specialist 的输出质量
特征：
职责：解析用户意图，分发给正确的 specialist
特征：
message:
from: agent_id # 发件人
to: agent_id # 收件人
type: task | report | question | escalate | approve
priority: high | normal | low
task:
id: uuid # 任务唯一ID
description: string # 任务描述
context: [] # 关联上下文引用
deadline: timestamp # 可选截止时间
status:
state: pending | in_progress | blocked | done | escalate
blockers: [] # 阻塞原因
result: {} # 结果（完成时填写）
metadata:
confidence: 0-1 # 置信度
requires_approval: boolean
| Type | 方向 | 说明 |
|---|---|---|
task | Master → Specialist | 分配任务 |
report | Specialist → Master | 任务完成报告 |
question | Specialist ↔ Master | 边界模糊，需确认 |
escalate | Specialist → Master | 超出能力，请求升级 |
approve | Critic → Master | 审查结论 |
PENDING → IN_PROGRESS → [BLOCKED → IN_PROGRESS] → DONE
↓
ESCALATED → IN_PROGRESS / DONE
状态转移规则：
pending → in_progress：Router 确认后立即转移in_progress → blocked：遇到障碍（缺信息/缺权限）时转移，blocker 写入 messageblocked → in_progress：Master 解除障碍后恢复in_progress → escalated：Specialist 判断超出边界时转移any → done：Critic 批准 or Master 兜底完成每个 Agent 必须遵守的记忆写入规则：
| 系统 | 存储内容 | 写入时机 |
|---|---|---|
| MemPalace | 团队决策、系统改动、长期事实、跨 agent 知识 | 重要信息确认后立即写 |
| Cortex (MemClaw) | 会话偏好、临时对话记忆、用户画像 | 每次对话后增量写入 |
每次任务开始前必须执行双重召回：
1. cortex_search — MemClaw 向量检索
2. mempalace__mempalace_search — MemPalace 结构化检索
跨 Agent 协作时必须同时跑：
使用飞书 channel 时，必须指定 target 参数：
| 目标 Agent | target 参数 |
|---|---|
| Master（小语） | user:ou_<master_open_id> |
| Specialist（小美/大龙） | user:ou_<specialist_open_id> |
| 阿辉（用户） | 走 channel 默认路由 |
禁止：accountId: default — 会导致 400 错误
# my-team.yaml
team:
name: 我的团队
agents:
- id: master
name: 小语
role: master
channel: feishu
target: user:ou_xxx
- id: specialist-ops
name: 大龙
role: specialist
domain: [ops, security, diagnosis]
channel: feishu
target: user:ou_yyy
- id: specialist-content
name: 小美
role: specialist
domain: [news, creative, design]
channel: feishu
target: user:ou_zzz
用户输入 → Router 解析意图 → Master 分配 → Specialist 执行 → Critic 审查 → Master 汇总
所有任务状态写入 MemPalace：
# Task state stored in MemPalace
task_id: "uuid-xxx"
state: "in_progress"
owner: "specialist-ops"
created_at: timestamp
updated_at: timestamp
history: [state_transitions]
真实生产配置（小语/小美/大龙）：
| Agent | 角色 | 飞书 ID | 职责边界 |
|---|---|---|---|
| 🌸 小语 | Master | ou_7d0630bcf6e7fb2c35c2033be5e23722 | 调度、审查、路由、兜底 |
| 🐉 大龙 | Specialist | ou_6a94e9d0e8eb6112626bdacefd0f3438 | 运维、诊断、系统检查、安全审计 |
| 🌷 小美 | Specialist | ou_96c1622be4c5efd17929c0d2ba5d3f99 | 新闻资讯、市场动态、图片、音乐、视频、文案 |
agent-swarm/
├── SKILL.md # This file
├── configs/
│ ├── team.yaml # 团队配置模板
│ └── roles/
│ ├── master.yaml # Master 角色定义
│ ├── specialist.yaml # Specialist 角色定义
│ ├── critic.yaml # Critic 角色定义
│ └── router.yaml # Router 角色定义
└── examples/
├── workflow-basic.yaml # 基础工作流示例
└── workflow-research.yaml # 研究任务工作流示例
Build high-performing OpenClaw agents end-to-end. Use when you want to design a new agent (persona + operating rules) and generate the required OpenClaw workspace files (SOUL.md, IDENTITY.md, AGENTS.md, USER.md, HEARTBEAT.md, optional MEMORY.md + memory/YYYY-MM-DD.md). Also use to iterate on an existing agent's behavior, guardrails, autonomy model, heartbeat plan, and skill roster.
brain v1.1.7 用户版（免费）：让AI Agent拥有类大脑决策架构。置信度触发、双保险记忆、经验胶囊复用。开箱即用，3片段内存池，2并发上限。
Orchestrate multi-agent teams with defined roles, task lifecycles, handoff protocols, and review workflows. Use when: (1) Setting up a team of 2+ agents with different specializations, (2) Defining task routing and lifecycle (inbox → spec → build → review → done), (3) Creating handoff protocols between agents, (4) Establishing review and quality gates, (5) Managing async communication and artifact sharing between agents.
Recommend curated prompts from a 10,000+ real-world image generation prompt library. Works with ANY AI image model — Nano Banana Pro, Nano Banana 2, Seedream 5.0, GPT Image 1.5, Midjourney, DALL-E 3, Flux, Stable Diffusion, and more. Use this skill when users want to: - Find proven image generation prompts (any model) - Get prompt inspiration for portraits, products, social media, posters, etc. - Create illustrations for articles, videos, podcasts, or marketing content - Browse categorized prompt templates with sample images - Translate and understand prompt techniques
A股量化数据分析工具，基于AkShare库获取A股行情、财务数据、板块信息等。用于回答关于A股股票查询、行情数据、财务分析、选股等问题。
Personalized news briefings from your chosen sources, delivered morning or evening, with voice option and smart filtering.