首页龙虾技能列表 › Eo Ability Multi Expert

Eo Ability Multi Expert

v1.0.0

多专家编排能力(Multi-Expert Orchestrator),协调多个专家并行/串行工作,结果自动汇流

0· 54·0 当前·0 累计
by @467718584·MIT-0
下载技能包
License
MIT-0
最后更新
2026/4/8
安全扫描
VirusTotal
无害
查看报告
OpenClaw
安全
high confidence
Instruction-only orchestration skill whose declared capabilities match its minimal footprint—no installs, no env vars, and no surprising actions in the SKILL.md.
评估建议
This skill is instruction-only and internally consistent, but before installing or enabling it: 1) Verify the skill's source and trustworthiness (homepage/source unknown). 2) If you pair it with an actual EO plugin or eo-workflow-* integrations, inspect those implementations—they may request network access or credentials not present here. 3) Watch for any runtime prompts that ask for API keys or system credentials and avoid granting unrelated secrets. 4) Prefer least-privilege for any EO/session...
详细分析 ▾
用途与能力
The name/description (multi-expert orchestrator) align with the instructions: splitting tasks, scheduling experts, merging results, checkpointing. The skill declares no extra binaries, env vars, or config paths, which is proportionate to an instruction-only orchestrator.
指令范围
SKILL.md contains orchestration logic and usage examples (/dream, Team Manager). It does not instruct the agent to read local files, system paths, or unrelated environment variables, nor does it instruct exfiltration or calls to unexpected external endpoints. It does reference EO sessions API and eo-workflow-* callers as integration points, which is expected for an orchestrator.
安装机制
No install spec and no code files are present (instruction-only). This minimizes on-disk execution and supply-chain risk.
凭证需求
The skill requests no environment variables or credentials. This is proportionate to the stated purpose. Note: actual integrations with an EO plugin or sessions API (mentioned in the doc) could require credentials in a concrete implementation—those are not present here.
持久化与权限
always is false and default autonomous invocation is allowed (normal). The skill does not request permanent presence or claim to modify other skills or system-wide settings.
安全有层次,运行前请审查代码。

License

MIT-0

可自由使用、修改和再分发,无需署名。

运行时依赖

无特殊依赖

版本

latestv1.0.02026/4/8

Initial release of eo-ability-multi-expert — Multi-Expert Orchestrator capability. - Supports orchestrating multiple experts in parallel or sequential execution. - Automatically merges expert outputs and manages result dependencies. - Handles expert result conflicts and supports checkpointing after configurable intervals. - Can run independently or as a core EO plugin, adapting to presence or absence of EO. - Exposes an interface for specifying experts, execution strategy, checkpoints, and timeouts.

● 无害

安装命令 点击复制

官方npx clawhub@latest install eo-ability-multi-expert
镜像加速npx clawhub@latest install eo-ability-multi-expert --registry https://cn.clawhub-mirror.com

技能文档

多专家编排能力 (Multi-Expert Orchestrator) - 协调多个专家并行/串行工作

一句话介绍

EO核心编排能力,多专家并行启动、结果自动汇流、依赖管理、冲突解决。

核心功能

  • 多专家并行: 多个专家同时启动
  • 结果汇流: 专家输出自动合并
  • 依赖管理: 专家执行顺序控制
  • 冲突解决: 多专家结果冲突处理
  • Checkpoint: 每N个专家后自动Checkpoint

使用方法

# 启动多专家协作
/dream "开发博客系统"

// 或通过 Team Manager 创建团队: 博客开发团队 添加专家: Architect, Backend, Frontend, QA 启动协作: parallel

与EO插件的协同

  • 被所有 eo-workflow- 调用
  • 是 EO 的核心差异化能力

独立运行模式(有EO vs 无EO)

模式能力
有EO插件真实多Agent并行、sessions API调用、结果汇流
无插件(基础)LLM模拟多专家、串行执行

执行流程

用户: /dream "开发博客系统"
           │
           ▼
┌─────────────────────────────────────┐
│  Multi-Expert Orchestrator           │
│  解析任务 → 分解 → 调度             │
└─────────────────┬───────────────────┘
                  │
    ┌─────────────┼─────────────┐
    ▼             ▼             ▼
┌────────┐  ┌────────┐  ┌────────┐
│Architect│  │Backend │  │Frontend│
│  Agent  │  │  Agent │  │  Agent │
└────┬────┘  └────┬────┘  └────┬────┘
     │            │            │
     └────────────┼────────────┘
                  ▼
         ┌──────────────┐
         │  Result Merge │
         │  + Checkpoint │
         └──────────────┘

Interface

Input

interface MultiExpertInput {
  experts: ExpertRequest[]
  strategy: 'parallel' | 'sequential' | 'pipeline'
  checkpointAfter?: number      // 每 N 个专家后 Checkpoint
  timeout?: number              // 超时时间(ms)
}

Output

interface MultiExpertOutput {
  results: ExpertResult[]
  checkpoints: CheckpointResult[]
  summary: string
  duration: number
}

🦞⚙️ 钢铁龙虾军团*

数据来源:ClawHub ↗ · 中文优化:龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制

免费技能或插件可能存在安全风险,如需更匹配、更安全的方案,建议联系付费定制

了解定制服务