← Back to Skills Marketplace
384
Downloads
1
Stars
0
Active Installs
6
Versions
Install in OpenClaw
/install agent-creator-skill
Description
Automatically create a new OpenClaw agent, translate its name, and initialize its persona/system prompt based on user requests.
README (SKILL.md)
\r \r
Agent Creator Skill\r
\r When a user asks you to create a new agent, assistant, or proxy (e.g., "创建一个Java高级开发的代理", "Help me build a product manager agent"), you MUST use this skill to accomplish the task.\r \r
SOP (Standard Operating Procedure):\r
\r
- Translation: Translate the requested role into a short English snake_case string. This will be the
\x3Cagent_name_en>.\r - Language Matching: Detect the primary language of the user's request. Use that language for both
\x3Cagent_display_name>and\x3Cidentity_prompt>, unless the user explicitly asks for another language. - Persona Generation: Generate a professional, highly detailed system prompt for this specific role. This will be the
\x3Cidentity_prompt>.\r - Execution: Choose the script that matches the user's operating system:
- On Linux, macOS, WSL, or Git Bash:
bash {baseDir}/create_agent.sh "\x3Cagent_name_en>" "\x3Cagent_display_name>" "\x3Cidentity_prompt>"- On native Windows PowerShell:
powershell -ExecutionPolicy Bypass -File "{baseDir}/create_agent.ps1" -AgentId "\x3Cagent_name_en>" -DisplayName "\x3Cagent_display_name>" -IdentityPrompt "\x3Cidentity_prompt>"
Usage Guidance
This skill's code largely matches its stated purpose, but review these before installing: 1) The project documents a Windows PowerShell script (create_agent.ps1) but that file is not included — on Windows the documented command would fail or cause the agent to try other behaviors. 2) The bash script edits ~/.openclaw/openclaw.json and sends the provided persona to the new agent; back up openclaw.json before first run and inspect the persona content you will inject. 3) The README suggests having your main agent pull code from GitHub and execute it; only do that from repositories you trust, because pulling arbitrary code to run on your system is how supply-chain compromises occur. 4) Verify the openclaw CLI binary is the legitimate one on PATH (this script invokes it directly). If you decide to use this skill: run create_agent.sh manually first (inspect and run in a shell) rather than allowing an automated installer to fetch and execute the repo, and ensure you understand the agent names and persona text that will be written to your OpenClaw config.
Capability Analysis
Type: OpenClaw Skill
Name: agent-creator-skill
Version: 1.0.6
The skill provides automated agent creation by executing shell/PowerShell scripts that modify the core application configuration file (~/.openclaw/openclaw.json) and execute CLI commands. While these actions are consistent with the stated purpose in SKILL.md and README.md, the use of 'ExecutionPolicy Bypass' in PowerShell and direct file system modification via inline Python scripts in create_agent.sh constitute high-risk behaviors that could be leveraged for unauthorized configuration changes if the agent is manipulated.
Capability Assessment
Purpose & Capability
The skill's name/description align with its behavior: the shell script creates an agent via the openclaw CLI, updates ~/.openclaw/openclaw.json, and injects a persona. Those actions are expected for an 'agent-creator' skill. Minor mismatch: documentation and SKILL.md reference a Windows PowerShell script (create_agent.ps1) and a {baseDir} placeholder, but no PowerShell script is present in the file manifest.
Instruction Scope
SKILL.md prescribes translating names, generating a persona, then running a system script. The provided bash script performs only the described tasks and touches only OpenClaw-related paths (~/.openclaw). It does modify openclaw.json and sends the persona text to the agent (openclaw agent --message). This is within scope, but the SKILL.md's strong imperative ('MUST use this skill') and references to a missing create_agent.ps1 and {baseDir} introduce ambiguity in Windows usage and in how the main agent should invoke the script.
Install Mechanism
There is no formal install spec (lowest install risk). The README suggests the main agent can pull the repo from GitHub to install — that step (pulling and executing code from an external repo) is not part of the skill bundle itself and carries the usual trust risk: if you instruct your main agent to fetch an arbitrary GitHub repo, that code will run on your system. The package itself contains only a local bash script and docs; no third-party downloads or obfuscated installers are present.
Credentials
The skill does not request environment variables, credentials, or unrelated system config. The script uses only $HOME and the OpenClaw config file (~/.openclaw/openclaw.json). It does not attempt to read or exfiltrate other system files or external credentials.
Persistence & Privilege
The skill is not marked always:true and does not request elevated platform privileges. It writes to OpenClaw-specific locations (workspace directory and openclaw.json) and invokes the openclaw CLI — these are appropriate for an agent-creation tool, but they are persistent changes to your OpenClaw configuration, so you should be comfortable with automatic edits to ~/.openclaw/openclaw.json and the automatic injection of persona text.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install agent-creator-skill - After installation, invoke the skill by name or use
/agent-creator-skill - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.6
- Added support for language detection: agent display names and persona prompts now match the user's primary request language unless specified otherwise.
- Changed parameter agent_name_cn to agent_display_name, allowing friendly names in any language.
- Updated example values and descriptions to reflect multi-language support.
- Updated SOP: now includes language detection and cross-platform execution scripts for Linux/macOS/WSL/Git Bash (bash) and Windows PowerShell.
v1.0.5
- Added a Chinese-language README file (README.zh-CN.md).
- Updated sample user requests in SKILL.md to include an English example for broader clarity.
- No changes to logic or functionality.
v1.0.3
- Added support for specifying a friendly Chinese display name (agent_name_cn) for agents.
- agent_name_en description clarified and example updated.
- identity_prompt parameter description made more general.
- Skill script invocation updated to include the new agent_name_cn parameter.
v1.0.2
No functional changes in this version.
- No file changes detected.
- Documentation, parameters, and usage remain the same.
v1.0.1
- No code or documentation changes detected.
- Version updated to 1.0.1; contents and instructions remain unchanged.
v1.0.0
Agent Creator Skill v1.0.0 – Initial Release
- Introduces a skill for automatic creation of OpenClaw agents based on user requests.
- Translates requested agent roles into English snake_case names.
- Generates a detailed Chinese system prompt/persona for each agent.
- Provides clear SOP for translating, generating persona, and executing agent creation via shell script.
Metadata
Frequently Asked Questions
What is agent-creator-skill?
Automatically create a new OpenClaw agent, translate its name, and initialize its persona/system prompt based on user requests. It is an AI Agent Skill for Claude Code / OpenClaw, with 384 downloads so far.
How do I install agent-creator-skill?
Run "/install agent-creator-skill" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is agent-creator-skill free?
Yes, agent-creator-skill is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does agent-creator-skill support?
agent-creator-skill is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created agent-creator-skill?
It is built and maintained by freesaber (@freesaber); the current version is v1.0.6.
More Skills