← Back to Skills Marketplace
Agent Pool
by
renxm78-creator
· GitHub ↗
· v1.0.0
· MIT-0
60
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install agent-pool
Description
全局Agent池。收录所有业务Agent的能力描述、工具依赖、记忆路径、工作区、触发条件。业务系统按需调用Agent，不绑定、不预分配。触发：查询Agent清单、编排新流程、新增Agent接入。
README (SKILL.md)
全局Agent池
核心原则
Agent（Agent）= 独立业务能力单元
Agent池（Agent Pool）= Agent注册表 + 能力定义
系统（System）= 流程编排层，按需调用Agent，不绑定
核心关系：
- Agent自己声明工具（工具池提供能力支撑）
- 系统只做流程编排（调用哪个Agent，不预分配）
- Agent调用产生的数据 → 写入发起调用的系统目录
Agent注册表
所有Agent按职能分为三层：
执行层Agent（直接产出业务价值）
| Agent代号 | 能力描述 | 工具依赖 | 记忆路径 | 工作区 | 触发条件 |
|---|---|---|---|---|---|
| 虾调查 | 爬取/清洗/去重/归档 → 结果写入本地知识库，维护分类/更新/版本/来源 | tavily / serper / reddit / linkedin / weixin / 视频摘要 / YouTube字幕 / baidu-ai-search | agents/research-agent/MEMORY.md |
agents/research-agent/ |
CMO调度 / 系统流程调用 |
| 虾写作 | 文案/邮件/落地页/技术文章 | read / write / edit | agents/writing-agent/MEMORY.md |
agents/writing-agent/ |
CMO调度 / 内容链路调用 |
| 虾审核 | 内容质量审核/7维度评分/否决机制 | read / write | agents/review-agent/MEMORY.md |
agents/review-agent/ |
虾写作完成后触发 |
策略层Agent（调度与规划）
| Agent代号 | 能力描述 | 工具依赖 | 记忆路径 | 工作区 | 触发条件 |
|---|---|---|---|---|---|
| 虾调度 | 分发策略/渠道组合/节奏控制/内容筛选 | read / write | agents/distribution-commander/MEMORY.md |
agents/distribution-commander/ |
CMO调度 / 内容审核通过后触发 |
| 虾规划 | 内容策略/关键词矩阵/SEO规划 | read / write | agents/content-strategist/MEMORY.md |
agents/content-strategist/ |
CMO调度 |
| 虾优化 | 效果分析/四象限分类/策略反馈 | read / write | agents/growth-analyst/MEMORY.md |
agents/growth-analyst/ |
内容发布后数据触发 |
触达层Agent（直接客户接触）
| Agent代号 | 能力描述 | 工具依赖 | 记忆路径 | 工作区 | 触发条件 |
|---|---|---|---|---|---|
| LinkedIn Agent | B2B触达/帖子发布/评论互动 | 临时声明 | agents/linkedin-agent/MEMORY.md |
agents/linkedin-agent/ |
虾调度调用 |
| Email Agent | 开发信/跟进邮件/序列发送 | 临时声明 | agents/email-agent/MEMORY.md |
agents/email-agent/ |
虾调度调用 |
| Video Agent | 信任增强视频/案例展示 | 临时声明 | agents/video-agent/MEMORY.md |
agents/video-agent/ |
虾调度调用 |
| Platform Agent | B2B平台询盘获客/A类客户筛选 | 临时声明 | agents/platform-agent/MEMORY.md |
agents/platform-agent/ |
虾调度调用 |
| Community Agent | 社区口碑/问答外链/长尾流量 | 临时声明 | agents/community-agent/MEMORY.md |
agents/community-agent/ |
虾调度调用 |
专项Agent
| Agent代号 | 能力描述 | 工具依赖 | 记忆路径 | 工作区 | 触发条件 |
|---|---|---|---|---|---|
| 虾展会 | 展会全流程（沙特FMF等）/展位设计/现场执行 | 临时声明 | agents/expo-operator/MEMORY.md |
agents/expo-operator/ |
CMO调度 / 展会筹备期 |
| Brand & Website | 官网优化/产品页/案例系统/信任承载 | 临时声明 | agents/brand-website-agent/MEMORY.md |
agents/brand-website-agent/ |
CMO直接调度 |
| 虾排名 | SEO架构/关键词策略/技术SEO/Schema | 临时声明 | agents/seo-architect/MEMORY.md |
agents/seo-architect/ |
CMO调度 |
Agent工具声明规则
- 虾调查：7个专属工具自动声明（见上方表格）
- 其他所有Agent：在任务指令中临时声明所需工具，不预持有
新增Agent接入流程
Step 1: 在本文件「Agent注册表」中新增Agent条目
Step 2: 创建/更新对应工作区的 MEMORY.md
Step 3: 系统在流程编排中按需调用（无需修改本文件）
参考文件
- 工具池（查询工具清单）：
skills/tool-pool/SKILL.md - 获客系统流程编排：
system-lead-generation/MEMORY.md - 早报系统流程编排：
system-morning-report/MEMORY.md - 新系统接入模板：
skills/agent-pool/references/system-template.md
Usage Guidance
Review this before installing in any workspace with sensitive data or autonomous agents. Use it only if you are comfortable with a global agent registry that documents memory paths and may guide agents to write into shared knowledge-base and system directories. Add explicit approval gates, per-agent caller rules, scoped write paths, and retention/provenance rules before using it for real orchestration.
Capability Assessment
Purpose & Capability
The skill's stated purpose matches its contents: it defines a global Agent Pool, agent capabilities, tool dependencies, memory paths, workspaces, and trigger conditions. The concern is not deception, but the breadth of registry and orchestration authority.
Instruction Scope
Activation phrases such as querying the agent list, orchestrating new workflows, onboarding agents, CMO scheduling, and system workflow calls are broad and could route ordinary coordination tasks into this registry without clear authorization boundaries.
Install Mechanism
The artifact contains markdown documentation and reference files only, with no executable scripts, package installs, startup hooks, or hidden install behavior found.
Credentials
For a global agent registry, exposing memory paths, workspaces, tool dependencies, trigger rules, and shared knowledge-base behavior is powerful and only partially scoped; external collection and outbound touch agents are listed without concrete caller or approval constraints.
Persistence & Privilege
The artifact instructs agents to write outputs into system directories and a shared knowledge base readable by all systems, but it does not define consent, data classification, provenance review, retention, or write allowlists.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install agent-pool - After installation, invoke the skill by name or use
/agent-pool - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
agent-pool v1.0.0
- Initial release introducing a global Agent Pool for unified business Agent registration and capability management.
- Provides structured registry tables for Execution, Strategy, Outreach, and Specialized Agents with fields for capability, tool dependencies, memory paths, workspace, and trigger conditions.
- Defines agent declaration, tool dependency rules, and a clear onboarding workflow for new Agents.
- Supports dynamic, on-demand Agent invocation in business systems without pre-allocation or binding.
- References related system flow templates and external tool pools for integration.
Metadata
Frequently Asked Questions
What is Agent Pool?
全局Agent池。收录所有业务Agent的能力描述、工具依赖、记忆路径、工作区、触发条件。业务系统按需调用Agent，不绑定、不预分配。触发：查询Agent清单、编排新流程、新增Agent接入。 It is an AI Agent Skill for Claude Code / OpenClaw, with 60 downloads so far.
How do I install Agent Pool?
Run "/install agent-pool" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Agent Pool free?
Yes, Agent Pool is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Agent Pool support?
Agent Pool is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Agent Pool?
It is built and maintained by renxm78-creator (@renxm78-creator); the current version is v1.0.0.
More Skills