← Back to Skills Marketplace
weiransoft

multi-agent-team

by Wang Wei · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
280
Downloads
1
Stars
2
Active Installs
1
Versions
Install in OpenClaw
/install multi-agent-team
Description
基于任务类型动态调度到合适的智能体角色(架构师、产品经理、测试专家、独立开发者)。支持多智能体协作、共识机制和完整项目生命周期管理。支持中英文双语。
Usage Guidance
This skill appears to do what it claims (role dispatch, project analysis, code-map, spec tools) and does not request credentials or perform external installs. Before installing or running it: 1) Review the bundled Python files (trae_agent_dispatch.py, project_understanding.py, code_map_generator.py, agent_loop_controller.py, spec_tools.py, etc.) for any network operations (requests, sockets, urllib, subprocess calls) or hardcoded endpoints/credentials. 2) Confirm the script paths referenced in SKILL.md exist in the package or adjust commands (SKILL.md uses scripts/*.py while files appear at repo root). 3) Run the skill in an isolated environment (container or VM) first, especially if you will point it at repositories or directories containing secrets or CI credentials. 4) Inspect the .trae-multi-agent progress files after runs; they store task lists, code-change history and verification results locally. 5) If you need strict control, disable autonomous invocation or review logs/outputs frequently. If you want, I can scan the actual Python sources for network calls and suspicious behaviors — upload the full contents of the key .py files and I will highlight any risky code paths.
Capability Analysis
Type: OpenClaw Skill Name: multi-agent-team Version: 1.0.0 The bundle implements a sophisticated multi-agent orchestration framework for software development, defining roles such as Architect and Solo Coder. It includes several Python scripts that perform high-risk operations necessary for its stated purpose: 'code_map_generator.py' and 'project_understanding.py' perform recursive filesystem scans to analyze project structure and dependencies, while 'agent_loop_controller.py' uses the subprocess module to execute internal dispatch scripts. Although these capabilities are aligned with the goal of project analysis and task management, the broad filesystem access and script execution represent a significant attack surface. No evidence of intentional malice, data exfiltration, or shell injection vulnerabilities was found.
Capability Assessment
Purpose & Capability
The name/description (multi-agent team, role dispatch, project lifecycle) match the provided Python scripts and extensive documentation (dispatcher, code-map, project-understanding, spec tools, agent loop). Small mismatch: SKILL.md examples call scripts/*.py, but the repository manifest lists the Python modules at repo root (e.g., agent_loop_controller.py, trae_agent_dispatch.py) — ensure the expected paths exist in your runtime environment or the entrypoint wrappers are present.
Instruction Scope
Runtime instructions tell the agent/user to run local Python scripts that read project directories and persist progress (e.g., .trae-multi-agent/progress.md). Reading project files and writing local progress is expected for 'project understanding' and 'code map' features, but that means the skill will access arbitrary files under any path you point it at — which can include secrets or private data. There are no instructions in SKILL.md to send data to external services, but you should still review the code for network calls before running.
Install Mechanism
No install spec is provided (instruction-only install), which is the lowest-risk model. The skill bundle includes Python source files that will be executed by the agent; nothing in the metadata indicates downloads from external URLs or arbitrary installers.
Credentials
The skill declares no required environment variables, credentials, or config paths. That aligns with its stated local analysis/dispatch functionality. There are no unexplained secret requests in the manifest or SKILL.md.
Persistence & Privilege
always:false (good). The skill persists progress to a local folder (.trae-multi-agent) and supports automatic continuation when thinking limits are reached. Autonomous invocation is allowed by platform default — combine that with the automatic-continue behavior could make the skill run longer or repeatedly without further prompts. This is expected for an automation-oriented dispatcher, but worth knowing.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install multi-agent-team
  3. After installation, invoke the skill by name or use /multi-agent-team
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
基于任务类型动态调度到合适的智能体角色(架构师、产品经理、测试专家、独立开发者)。支持多智能体协作、共识机制和完整项目生命周期管理。支持中英文双语。Dynamically dispatches to appropriate agent roles (Architect, Product Manager, Test Expert, Solo Coder) based on task type. Supports multi-agent collaboration, consensus mechanism, complete project lifecycle management, specification-driven development, code map generation, and project understanding. Supports Chinese-English bilingual.
Metadata
Slug multi-agent-team
Version 1.0.0
License MIT-0
All-time Installs 2
Active Installs 2
Total Versions 1
Frequently Asked Questions

What is multi-agent-team?

基于任务类型动态调度到合适的智能体角色(架构师、产品经理、测试专家、独立开发者)。支持多智能体协作、共识机制和完整项目生命周期管理。支持中英文双语。 It is an AI Agent Skill for Claude Code / OpenClaw, with 280 downloads so far.

How do I install multi-agent-team?

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

Is multi-agent-team free?

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

Which platforms does multi-agent-team support?

multi-agent-team is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created multi-agent-team?

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

💬 Comments