← Back to Skills Marketplace
534422530

Collaborative Agent

by 534422530 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
45
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install collaborative-agent
Description
ARC多智能体协作:规划/执行/审查/总结/研究/批评,迭代评审
README (SKILL.md)

Skill: collaborative-agent

ARC 式多智能体协作框架

归属管理

  • 归属核心:think-expand(想核心)
  • 上级负责:复杂决策任务由think-expand分解后委托协作框架执行

功能范围

  • 多角色协作:规划者/执行者/审查者/总结者/研究者/批评者
  • 迭代式同行评审:产出→评审→反馈→优化
  • 任务分解:复杂任务自动拆分为子任务
  • 依赖管理:子任务按依赖顺序执行
  • 结果综合:多步结果整合为结构化报告

原理

源自 ARC (AutoResearchClaw 2026) 研究:

  • 多智能体同行评审协作
  • 自我强化循环
  • 集成已有能力:ScriptDB / VSAEngine / SkillEvolver / CognitiveReasoner

用法

import sys; sys.path.insert(0, r'D:\coze-local\db')
from collaborative_agent import ARCWorkflow, AgentRole

# 创建协作工作流
wf = ARCWorkflow()

# 添加任务(自动分配角色)
wf.add_task("研究Transformer优化方案", role="researcher")
wf.add_task("制定实现计划", role="planner", depends_on=[0])
wf.add_task("实现并测试", role="executor", depends_on=[1])
wf.add_task("审查代码质量", role="reviewer", depends_on=[2])
wf.add_task("生成总结报告", role="summarizer", depends_on=[3])

# 执行所有任务
results = wf.run_all()

# 获取最终报告
report = wf.get_report()

触发场景

  • 复杂问题需要分步骤处理(规划→执行→审查→总结)
  • 需要多角度分析的研究任务
  • 需要质量审查的代码/内容生成
  • 需要从多来源综合信息的任务

集成

  • 内部自动调用 capability_executor.detect_and_execute() 执行已注册能力
  • 使用 CognitiveReasoner 提供语言/数值反馈
  • 使用 SkillEvolver 记录失败并进化

Base directory: file:///C:/Users/pc/.config/opencode/skills/collaborative-agent \r \r

B站学习\r

学习时间: 2026-06-01 20:57\r \r

  • 安逸Ai丶: Agent中为什么需要 tracing 和 observability?\r
    • 关键词: Agent中为什么需要, tracing, observability\r
  • 银色海豚KK: OpenClaw v2026.5.28 更新快报 | ClawPDF Claude Opus 4.8 Agent协作\r
    • 关键词: OpenClaw, v2026, 28, 更新快报, ClawPDF\r \r

B站学习\r

学习时间: 2026-06-01 21:01\r \r

  • 安逸Ai丶: Agent中为什么需要 tracing 和 observability?\r
  • 银色海豚KK: OpenClaw v2026.5.28 更新快报 | ClawPDF Claude Opus 4.8 Agent协作\r
  • 智能体老王: AI给自己写工具?字节团队出品 MUSE-Autoskill 让Agent实现技能自进化,推动skills成为可复用的知识资产\r \r

融合来源: collaborative-agent-1fe479\r

融合时间: 自动合并\r \r 学习时间: 2026-06-01 21:07\r

  • WJ_UPC: 20221216_王钧_Advances in Collaborative Neurodynamic Optimization\r
  • 树欲静心不止: 250908- Coding Agent实战心得:从工具到伙伴的协作之道 / Coding Agent Mastery: Collaborative Devel\r
  • UncleScto: 3D LMI Gocator with HANWHA collaborative robot\r

融合时间: 自动合并\r 学习时间: 2026-06-02 07:52\r

  • dailyrain: 3GC Collaborative and Creative Content Generation in Game Design\r
  • dailyrain: 3GC Collaborative and Creative Content Generation in Game Design\r
  • Tracy春雪: The Teachers' Room - Collaborative Writing 1 - Planning\r \r

B站学习 (第1轮)\r

学习时间: 2026-06-02 09:20\r \r

B站学习 (第2轮)\r

学习时间: 2026-06-02 09:33\r \r

Usage Guidance
Review this skill before installing if your environment has powerful registered capabilities, private data stores, or tools that can modify files, accounts, services, or installed skills. Prefer using it only where capability execution requires explicit approval and where SkillEvolver logging or behavior changes are understood and controllable.
Capability Assessment
Purpose & Capability
The stated purpose is multi-role planning, execution, review, research, and summarization, which fits the described workflow; however, automatically executing registered capabilities can become high-impact depending on what capabilities are available.
Instruction Scope
Trigger conditions cover broad complex, research, code/content generation, and synthesis tasks, and the integration notes say it internally calls `capability_executor.detect_and_execute()` without defining limits, dry-run behavior, approval checkpoints, or sensitive-task exclusions.
Install Mechanism
The package contains only a markdown skill file, no executable scripts, dependency declarations, installer, or bundled binary artifacts.
Credentials
The example imports code from a local Windows path and references local components such as ScriptDB, VSAEngine, SkillEvolver, and CognitiveReasoner; these are purpose-related but their permissions and data access are not described in the artifact.
Persistence & Privilege
The skill says it uses `SkillEvolver` to record failures and evolve, implying persistent behavior changes or logging, but it does not explain where data is stored, how users can inspect it, or how to disable or reverse it.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install collaborative-agent
  3. After installation, invoke the skill by name or use /collaborative-agent
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
collaborative-agent v1.0.0 - Initial release of an ARC-style multi-agent collaboration framework. - Supports multi-role collaboration: planner, executor, reviewer, summarizer, researcher, critic. - Features iterative peer review, automatic task decomposition, and dependency-based execution. - Integrates existing capabilities: ScriptDB, VSAEngine, SkillEvolver, CognitiveReasoner. - Provides structured and modular workflow for complex decision-making and multi-perspective research tasks.
Metadata
Slug collaborative-agent
Version 1.0.0
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 1
Frequently Asked Questions

What is Collaborative Agent?

ARC多智能体协作:规划/执行/审查/总结/研究/批评,迭代评审. It is an AI Agent Skill for Claude Code / OpenClaw, with 45 downloads so far.

How do I install Collaborative Agent?

Run "/install collaborative-agent" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Collaborative Agent free?

Yes, Collaborative Agent is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Collaborative Agent support?

Collaborative Agent is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Collaborative Agent?

It is built and maintained by 534422530 (@534422530); the current version is v1.0.0.

💬 Comments