← 返回 Skills 市场
congjianfei

Agent Collab Protocol

作者 congjianfei · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ 安全检测通过
49
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install agent-collab-protocol
功能描述
Standardized agent collaboration protocol for multi-agent systems. Defines L0/L1/L2 hierarchy, task contracts, error handling, agent templates, and a new-age...
使用说明 (SKILL.md)

Agent Collaboration Protocol

A complete protocol layer for building structured, multi-agent systems.
Provides the rules, templates, and operational procedures to ensure agents work together consistently.

What's Included

Component Description
AGENT_COLLAB_PROTOCOL.md Core protocol: L0/L1/L2 hierarchy, JSON task contracts, \x3Cagent-response> format, error handling, spawn vs send patterns
OPERATIONS.md Operational guide: how to add/remove domains, file ownership matrix, protocol change propagation
NEW_AGENT_CHECKLIST.md 8-step mandatory checklist for creating new sub-agents (no skip, no placeholder)
TEMPLATE_AGENTS.md Standard AGENTS.md template with protocol implementation table
TEMPLATE_IDENTITY.md Identity card template (Agent ID, layer, upstream/downstream, model)
TEMPLATE_SOUL.md Values and red lines template (core belief, values, anti-patterns)
TEMPLATE_USER.md Service object template (upstream, downstream consumers, service standards)
scaffold-domain.sh Shell script to auto-generate a new domain workspace from templates

Quick Start

Install from ClawHub

npx clawhub install agent-collab-protocol

Or use locally

# Protocol files are in: skill/agent-collab-protocol/references/
# Copy to your protocol directory:
cp -r skill/agent-collab-protocol/references/ ~/.openclaw/protocol/

Create a new business domain

cd ~/.openclaw/protocol
bash scaffold-domain.sh fintech 金融产品经理 "fintech-analyzer:金融分析师"

Architecture Overview

Layer 0 (Orchestrator)          Identify domain → dispatch to Manager → verify
    ↓ (natural language)
Layer 1 (Manager)               Understand → decompose → dispatch → aggregate
    ↓ (JSON Task Package)
Layer 2 (Specialist)            Execute → deliver via \x3Cagent-response>

Key Protocol Rules

  • L0 dispatches to L1 Managers via natural language
  • L1 decomposes tasks into JSON Task packages for L2 Specialists
  • L2 executes and returns \x3Cagent-response> blocks
  • Tasks that fail: retry 1 time → downgrade → report as partial/failed
  • 3-layer maximum depth (L0 → L1 → L2)

Usage

For Framework Authors

Reference AGENT_COLLAB_PROTOCOL.md to implement protocol compliance in your agent runtime.

For Agent Developers

  1. Copy the relevant TEMPLATE_*.md files
  2. Fill in your agent's business-specific content
  3. Register in your domain configuration
  4. Follow NEW_AGENT_CHECKLIST.md for new agents

For Team Leads

Use NEW_AGENT_CHECKLIST.md as a mandatory gating process — each team member must check off all 8 steps before a new agent is considered complete.

Files

agent-collab-protocol/
├── SKILL.md                       ← This file
├── _meta.json                     ← Package metadata
└── references/
    ├── AGENT_COLLAB_PROTOCOL.md   ← Core protocol (413 lines)
    ├── OPERATIONS.md              ← Operational guide
    ├── NEW_AGENT_CHECKLIST.md     ← Mandatory checklist (156 lines, 64 checks)
    ├── TEMPLATE_AGENTS.md         ← AGENTS.md template
    ├── TEMPLATE_IDENTITY.md       ← Identity card template
    ├── TEMPLATE_SOUL.md           ← Values & red lines template
    ├── TEMPLATE_USER.md           ← Service object template
    └── scaffold-domain.sh         ← Domain scaffolding script

Changelog

Version Date Changes
1.0.0 2026-06-04 Initial public release
安全使用建议
Before installing or using it, review the scaffold script and deletion instructions, validate domain names so paths stay under ~/.openclaw, and replace the example systemPromptOverride wording if your agents must preserve a stricter system/developer/user priority model.
能力评估
Purpose & Capability
The artifacts match the stated purpose: protocol docs, agent templates, a checklist, and a scaffold script for OpenClaw multi-agent workspaces.
Instruction Scope
The generated configuration examples use systemPromptOverride and tell agents to follow AGENTS.md, including a phrase that user instructions have highest priority; users should adapt that to their own instruction hierarchy and safety model.
Install Mechanism
Installation is ordinary skill/package installation; the scaffold script is included as a reference tool and is not shown as automatically executed on install.
Credentials
The scaffold script writes under ~/.openclaw/workspace and prints manual openclaw.json changes, which is proportionate for an agent workspace scaffold, though the domain argument is not path-validated.
Persistence & Privilege
The protocol includes optional persistent sessions and session directories for subagents, which is disclosed and purpose-aligned; there is no hidden persistence, credential use, or exfiltration behavior found.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agent-collab-protocol
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agent-collab-protocol 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial public release of agent-collab-protocol. - Provides a standardized protocol for multi-agent collaboration, defining L0/L1/L2 hierarchy, task contracts, agent templates, and operational procedures. - Includes ready-to-use templates (AGENTS, identity, values, service object), an 8-step new agent checklist, and a shell script for scaffolding new domains. - Compatible with OpenClaw and works with any agent framework.
元数据
Slug agent-collab-protocol
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Agent Collab Protocol 是什么?

Standardized agent collaboration protocol for multi-agent systems. Defines L0/L1/L2 hierarchy, task contracts, error handling, agent templates, and a new-age... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 49 次。

如何安装 Agent Collab Protocol?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-collab-protocol」即可一键安装,无需额外配置。

Agent Collab Protocol 是免费的吗?

是的,Agent Collab Protocol 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Agent Collab Protocol 支持哪些平台?

Agent Collab Protocol 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Agent Collab Protocol?

由 congjianfei(@congjianfei)开发并维护,当前版本 v1.0.0。

💬 留言讨论