A powerful management system to deploy and orchestrate a team of specialized AI agents with unique identities and model configurations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-team
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-team 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 Team Manager?
agent-team is a sophisticated framework designed to manage sub-agents within the Openclaw Skills ecosystem. By assigning unique souls and specific LLM backends to different agents, it allows developers to build a collaborative environment where each agent specializes in a distinct domain like coding, data analysis, or deep research.
This skill bridges the gap between single-prompt interactions and complex, multi-agent workflows. It enables more nuanced and high-quality outputs by leveraging the strengths of various models—such as Qwen for coding and Gemini for research—all within a unified interface.
Agent Team Manager Use Cases
- Automating full-stack development cycles by chaining coder and reviewer agents.
- Generating high-quality technical documentation by passing raw code to a specialized writer agent.
- Conducting deep market research using specialized agents backed by high-reasoning models.
- Parallelizing complex tasks by spawning multiple agents simultaneously to handle different parts of a project.
How Agent Team Manager Works
- The system initializes by scanning the workspace directory for defined agent profiles containing identity and configuration files.
- Users can list available agents or inspect specific agent souls and capabilities via the CLI.
- When a task is assigned via the spawn command, the skill creates an isolated sub-agent session with the specified persona and model.
- For interactive needs, the chat mode establishes a persistent session that maintains the agent's character and context throughout the conversation.
- Results can be piped between agents to create complex automation chains or parallelized for efficiency.
Agent Team Manager Setup
To integrate new agents into your Openclaw Skills workspace, follow these steps:
# 1. Create the specific agent directory
mkdir -p ~/.openclaw/workspace/agents/<agent-name>
# 2. Define the agent identity in SOUL.md
touch ~/.openclaw/workspace/agents/<agent-name>/SOUL.md
# 3. Configure the model and capabilities in config.json
touch ~/.openclaw/workspace/agents/<agent-name>/config.json
Agent Team Manager Data Schema & Taxonomy
The skill organizes data within a structured workspace to ensure portability and clarity within Openclaw Skills:
| File | Type | Description |
|---|---|---|
SOUL.md |
Markdown | Defines the persona, personality, and expertise of the agent. |
config.json |
JSON | Specifies primary/fallback models and functional capabilities. |
agent-team.py |
Python | The core orchestration logic for managing agent lifecycles. |
Agent Team Manager Advanced Features
- Parallel Multi-Agent Execution: Launch multiple agents concurrently to significantly reduce task completion time.
- Sequential Chaining: Create pipelines where the output of a Coder agent is automatically audited by a Reviewer agent.
- Dynamic Model Switching: Configure primary and fallback models in the config.json to ensure task reliability.
- Customizable Souls: Deeply define agent behavior and communication styles using the SOUL.md specification framework.
SKILL.md
Loading
Related Openclaw Skills
Agent Reach
A comprehensive toolset providing AI agents with direct access to 13+ major internet platforms and social media channels.
last30days
A high-speed research engine that scrapes Reddit, X, and the web to synthesize real-time insights and generate precision AI prompts.
Deep Search
A sophisticated research skill that performs iterative search queries and LLM synthesis to provide exhaustive, multi-source analysis.
Codex Deep Search
A sophisticated deep-research skill that utilizes Codex CLI to perform multi-source synthesis and high-context analysis for complex queries.
Auto Agent Router
A sophisticated command-driven dispatcher that routes user prompts to specialized child agents for coding, writing, research, and analysis.
DingTalk Group Saver
A background utility that automatically archives DingTalk group metadata and IDs to persistent storage whenever the agent is @mentioned.
Featured*
