← Back to Skills Marketplace
141
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install apollo-coordinator
Description
复杂任务交给多个小助手同时做，最后帮你汇总结果。
README (SKILL.md)
Apollo 协调器
触发场景
当用户提到"多任务并行"、"分解任务"、"分配给多个Agent"、"Hub+Spoke"、"协调子Agent"时激活。
核心概念
Hub（协调者）： 负责任务分解、结果汇总、最终交付，不做具体执行
Spoke（执行者）： 负责具体执行，独立完成各自子任务
工作流程
用户输入复杂任务
↓
Hub 分解任务为 N 个独立子任务
↓
并行分配给 N 个 Spoke（独立Agent）
↓
每个 Spoke 执行并返回结果
↓
Hub 汇总结果，处理冲突
↓
最终交付给用户
分解原则
- 独立性：每个Spoke能独立完成，不依赖其他Spoke
- 粒度合适：一个Spoke的工作量控制在10-30分钟
- 格式统一：每个Spoke返回结构化结果，方便Hub汇总
推荐子任务数量
| 任务复杂度 | 子任务数量 |
|---|---|
| 简单任务 | 2个 |
| 中等任务 | 3-4个 |
| 复杂任务 | 最多5个 |
超时处理
- 单个子任务超时 → 标记"超时"，继续其他任务
- 汇总时说明哪些超时了
冲突处理
- 多个Spoke返回冲突信息 → Hub标注冲突，让用户判断
输出格式
任务分解（Hub给出）
【任务分解】
主任务：{用户原始任务}
分解为 {N} 个子任务：
任务1：{描述}
- 目标：{具体目标}
- 输出格式：{结构化格式}
结果返回（Spoke给出）
【任务 {N} 结果】
状态：成功/超时/失败
输出：
{结构化内容}
注意事项
- Hub本身不执行具体工作，只做协调
- 如果用户要求一个人完成，不要强行分解
- Spoke之间不直接通信，都通过Hub中转
Usage Guidance
This skill appears to do what it claims: it inspects the agent workspace and reports coordinator health, and writes a local state file. Before installing, review whether your OpenClaw workspace (/root/.openclaw/workspace) contains any secrets or sensitive files, since the script enumerates the skills directory and reads JSON files there. If you run agents under a different user or directory, confirm the paths are acceptable or adjust them. Also note SKILL.md doesn't explicitly document the bundled script, so if you want to limit file access, consider inspecting or sandboxing the script (or running it manually) before allowing autonomous agent invocation.
Capability Analysis
Type: OpenClaw Skill
Name: apollo-coordinator
Version: 2.0.0
The bundle provides a coordination framework for multi-agent task execution using a Hub-and-Spoke model. The SKILL.md defines clear logic for task decomposition and result aggregation, while the scripts/coordinator/coordinator-check.sh utility performs legitimate health checks on workflow states and skill availability within the workspace. No indicators of data exfiltration, malicious execution, or prompt injection were found.
Capability Assessment
Purpose & Capability
The skill claims to coordinate multiple sub-agents (Hub/Spoke). The included script implements a coordinator health/status check by reading workflow, gate, skills, and task-status files under the agent workspace. These file accesses and checks are consistent with a coordinator's purpose.
Instruction Scope
SKILL.md describes task decomposition and coordination but does not explicitly mention running the included shell script. The script reads agent workspace files (e.g., /root/.openclaw/workspace/.workflow, /root/.openclaw/workspace/skills, /root/.openclaw/workspace/.dream/task-state.json and /tmp lock files) and writes a local state file. That file I/O is relevant to coordination, but users should be aware the skill inspects other files under the workspace.
Install Mechanism
There is no install spec or external download — the skill is instruction-only with a bundled script. No network installs or third-party package pulls are present.
Credentials
The skill does not request environment variables or credentials. It does read and write files under /root/.openclaw/workspace and /tmp; this is proportionate to a coordinator but means it can observe other skills' workspace files, so verify no sensitive secrets are stored there.
Persistence & Privilege
always:false (normal). The script creates and writes a persistent state file at /root/.openclaw/workspace/.coordinator/state.json. This is confined to the OpenClaw workspace and does not modify other skills' configurations, but it does create persistent data on disk.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install apollo-coordinator - After installation, invoke the skill by name or use
/apollo-coordinator - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.0
v2.0.0: coordinator-check.sh脚本实现
v1.1.0
描述优化：一句话非技术描述
v1.0.0
Apollo-coordinator 1.0.0 — Initial release
- Introduces a multi-Agent task coordination framework for decomposing complex tasks into independent subtasks, distributed among multiple Agents for parallel execution and results aggregation.
- Defines "Hub" (coordinator) and "Spoke" (executor) roles and a clear workflow from task input to result delivery.
- Provides guidelines for subtask independence, suitable granularity, result structuring, and recommended subtask counts based on task complexity.
- Includes mechanisms for timeout and conflict handling, with specific output formats for both task breakdown and subtask results.
- Ensures that all agent communication is managed via the Hub, with no direct interaction between Spokes.
Metadata
Frequently Asked Questions
What is Apollo Coordinator?
复杂任务交给多个小助手同时做，最后帮你汇总结果。 It is an AI Agent Skill for Claude Code / OpenClaw, with 141 downloads so far.
How do I install Apollo Coordinator?
Run "/install apollo-coordinator" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Apollo Coordinator free?
Yes, Apollo Coordinator is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Apollo Coordinator support?
Apollo Coordinator is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Apollo Coordinator?
It is built and maintained by nic-yuan (@nic-yuan); the current version is v2.0.0.
More Skills