A standardization layer that defines strict hierarchical interaction patterns, structured task contracts, and scaffolding utilities for multi-agent systems.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-collab-protocol
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-collab-protocol using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Collaboration Protocol?
The Agent Collaboration Protocol offers a comprehensive blueprint for orchestrating complex, multi-tiered AI agent workflows. By defining a clear hierarchy from Layer 0 Orchestrators down to Layer 2 Specialists, it eliminates conversational drift and guarantees predictable outputs. This package is an essential asset within the Openclaw Skills ecosystem, standardizing multi-agent execution at scale.
Built on structured JSON task packages and structured response boundaries, this protocol gives developers and teams the frameworks required to build resilient, self-healing agent systems. Whether you are bootstrapping a new business domain or enforcing quality control across developmental teams, this system provides the operational guardrails needed for production-grade AI.
Agent Collaboration Protocol Use Cases
- Structured Multi-Agent Execution: Orchestrate multi-agent networks without chaotic cross-talk or command loops by enforcing a strict L0/L1/L2 hierarchy.
- Rapid Domain Bootstrapping: Spin up entirely new operational agent environments in seconds utilizing standardized automation scripts.
- Quality Control & Gating: Implement mandatory, 8-step verification checklists to audit agent behaviors, identities, and technical standards before shipping to production.
- Resilient Task Recovery: Define strict error handling boundaries and automatic degradation paths when deep tasks fail.
How Agent Collaboration Protocol Works
- Task Identification: The Layer 0 Orchestrator analyzes incoming requests, identifies the target domain, and dispatches instructions to a Layer 1 Manager using natural language.
- Decomposition & Package Dispatching: The Layer 1 Manager parses the natural language prompt, decomposes it into structured tasks, and distributes these JSON Task Packages to Layer 2 Specialists.
- Specialized Execution: Layer 2 Specialists execute their discrete, dedicated tasks and return results wrapped in clean, parseable
<agent-response>blocks. - Failover & Error Resolution: If an agent task fails, the protocol triggers a retry sequence. If the retry fails, it initiates a safe task degradation sequence and logs a partial failure report back up the hierarchy.
Agent Collaboration Protocol Setup
To get started with the Agent Collaboration Protocol, install it via ClawHub or copy the files directly to your local protocol store:
Installation
npx clawhub install agent-collab-protocol
Local Reference Setup
If you prefer manual setup, copy the reference files to your local configuration path:
cp -r skill/agent-collab-protocol/references/ ~/.openclaw/protocol/
Scaffolding a Domain
Create a standardized domain workspace with the automatic scaffolding utility:
cd ~/.openclaw/protocol
bash scaffold-domain.sh fintech 金融产品经理 "fintech-analyzer:金融分析师"
Agent Collaboration Protocol Data Schema & Taxonomy
The protocol establishes a standardized workspace directory layout ensuring predictable template paths and tracking metadata.
| Directory / File | Description |
|---|---|
_meta.json |
Manifest holding versioning details, tags, and skill taxonomy. |
references/AGENT_COLLAB_PROTOCOL.md |
Core specification defining task package syntax and response structures. |
references/OPERATIONS.md |
Deployment guide outlining ownership boundaries and protocol change rules. |
references/NEW_AGENT_CHECKLIST.md |
Mandatory auditing document containing 8 key stages and 64 strict validation checkpoints. |
references/TEMPLATE_AGENTS.md |
Base markdown workspace template specifying protocol implementation configurations. |
references/TEMPLATE_IDENTITY.md |
Configuration map standardizing upstream/downstream integrations and metadata. |
references/TEMPLATE_SOUL.md |
Behavioral constraint definitions including core beliefs, execution values, and anti-patterns. |
references/TEMPLATE_USER.md |
Integration contract specifying upstream requirements and downstream delivery rules. |
Agent Collaboration Protocol Advanced Features
- Automated Domain Scaffolding: Rapidly generate isolated agent environments using pre-configured shell scripts to minimize manual setup errors.
- Behavioral Boundaries & Anti-Patterns: Embed strict cognitive and operational constraints (using "soul" templates) to maintain agent alignment with business guidelines.
- Hierarchical Depth Restrictions: Enforce a strict maximum execution depth of 3 layers (L0 to L2) to eliminate runaway infinite execution loops.
- High-Fidelity Quality Gates: Leverage extensive operational checklists with 64 validation points designed for continuous integration pipeline validation.
SKILL.md
Loading
Related Openclaw Skills
li_patrol_2026
An intelligent multi-layered O&M automated inspection tool that auto-detects system services and computes health scores.
National Gaokao Volunteer Assistant (全国高考志愿填报助手)
An MCP-powered educational planning assistant designed to help Chinese high school students query historical admission scores, view enrollment plans, and calculate university entry probabilities.
Weekly Meeting Excel & Meixin Automation
Automatically locates, modifies, and sends weekly progress Excel sheets to Meixin work groups every Monday morning.
Lowcode Module Generator
The Lowcode Module Generator accelerates full-stack application development by converting module names and field schemas into production-ready Java backend and React frontend code.
My Work Profile Cross-Agent
A local, cross-agent context sync tool that dynamically profiles your development stack and business domains to deliver hyper-personalized AI code and advice.
vn-weijialiao (Micro Gia Liao - 18+ & Lore Integration)
An AI-powered novel enhancement skill designed to intelligently convert clean web novel chapters into 18+ variants and integrate deep lore seamlessly.
Featured*
