openclaw-multi-agent-wizard

OpenClaw Multi-Agent Wizard

中文说明

A beginner-friendly OpenClaw skill that guides users through multi-agent setup step by step, with safe defaults, simple explanations, and Feishu-focused onboarding.

Why This Repository Exists

Most multi-agent setup guides assume the user already understands agent IDs, bindings, routing, group mapping, and channel integration details.

This skill takes the opposite approach:

• start from a beginner-friendly question
• recommend the safest mode first
• generate starter agent profile files automatically
• explain each setup step in plain language
• verify before calling the setup complete

What It Helps You Do

• choose a beginner-friendly multi-agent mode
• create one or more OpenClaw agents
• generate starter profile files for each agent
• connect Feishu bots step by step
• configure safe routing defaults
• summarize the final setup in simple language

Supported Modes

1. Multi-bot, multi-agent

One bot maps to one agent.

Examples:

• work bot -> work assistant
• life bot -> life assistant

This is the recommended default for beginners.

2. Single-bot, multi-agent

One Feishu bot is shared, while different Feishu groups map to different agents.

Examples:

• product group -> product assistant
• engineering group -> engineering assistant

V1 supports group-based routing only for this mode.

3. A2A collaboration

One main agent responds publicly while one or more worker agents help in the background.

Examples:

• main assistant replies in Feishu
• data assistant gathers numbers
• writing assistant polishes the final answer

What Gets Generated

For each new agent workspace, the helper scripts can generate:

• IDENTITY.md
• SOUL.md
• AGENTS.md
• MEMORY.md
• TOOLS.md
• USER.md

This gives each agent a lightweight starter identity, role boundary, collaboration shape, and user-facing purpose.

Feishu Focus

This repository currently prioritizes Feishu because it is a practical onboarding path for beginner OpenClaw users.

The setup flow is intentionally broken into small steps:

1. create a Feishu app
2. enable bot capability
3. copy App ID and App Secret
4. finish event subscription
5. connect OpenClaw
6. bind the bot or group routing
7. verify the result

Project Principles

• preflight first
• one small step at a time
• safe defaults over advanced options
• minimal config edits
• verify before claiming success
• optimize for beginner success, not maximum feature coverage

Current Boundaries

• single-bot multi-agent supports Feishu group routing only
• private-chat routing is treated as advanced
• advanced runtime orchestration is explained carefully, not enabled by default
• same-group public multi-agent chatter is treated as experimental

Directory Structure

openclaw-multi-agent-wizard/
|-- SKILL.md
|-- README.md
|-- README.zh-CN.md
|-- LICENSE
|-- CONTRIBUTING.md
|-- CODE_OF_CONDUCT.md
|-- CONTRIBUTORS.md
|-- agents/
|   `-- openai.yaml
|-- assets/
|   |-- icon-small.svg
|   |-- icon-large.svg
|   `-- repo-preview.svg
|-- references/
|   |-- a2a-mode.md
|   |-- advanced-mode.md
|   |-- command-branches.md
|   |-- dialogue-scripts.md
|   |-- feishu-setup.md
|   |-- final-summary.md
|   |-- migration-existing-setup.md
|   |-- modes.md
|   |-- persona-templates.md
|   |-- preflight.md
|   |-- quick-start.md
|   |-- routing-basic.md
|   `-- troubleshooting.md
`-- scripts/
    |-- generate_agent_ids.py
    |-- render_setup_summary.py
    |-- suggest_persona_kind.py
    |-- write_identity_template.py
    `-- write_starter_profile.py

Key Files

• SKILL.md: main skill behavior and guardrails
• agents/openai.yaml: UI metadata
• references/quick-start.md: compact mental model
• references/feishu-setup.md: Feishu onboarding flow
• references/persona-templates.md: starter profile guidance
• references/a2a-mode.md: A2A collaboration notes
• scripts/write_starter_profile.py: starter profile bundle generator

Installation

Place this skill where OpenClaw can discover local skills.

Typical local skill location:

$OPENCLAW_HOME/skills/openclaw-multi-agent-wizard

If OPENCLAW_HOME is not set, a common default is:

~/.openclaw/skills/openclaw-multi-agent-wizard

Validation

Basic script validation:

python -m compileall scripts

If the local skill validator is available:

python /path/to/quick_validate.py /path/to/openclaw-multi-agent-wizard

Roadmap Ideas

• add onboarding paths beyond Feishu
• improve migration support for existing OpenClaw setups
• add deeper persona refinement for generated agents
• expand worker role categories for A2A collaboration

Community

• Contributing
• Code of Conduct
• Contributors

License

Released under the MIT License.