agent-creator-skill

Automatically create a new OpenClaw agent, translate its name, and initialize its persona/system prompt based on user requests.

Security Scan

The skill's code and runtime instructions are coherent with its stated purpose: it creates an OpenClaw agent, updates the local OpenClaw config, and injects a persona; it does not request unrelated credentials or external downloads.

Purpose & Capability

The name, description, SKILL.md, README, and the included create_agent.sh all consistently implement: translate a name, create an agent via the openclaw CLI, set the display name in ~/.openclaw/openclaw.json, and inject a persona message. There are no unrelated environment variables, binaries, or hidden capabilities requested.

Instruction Scope

The SKILL.md instructs the agent to run the bundled bash script, which reads and writes the user's OpenClaw config (~/.openclaw/openclaw.json), creates a workspace directory under $HOME, and invokes openclaw CLI commands to add the agent and send a message. Those actions are appropriate for the stated purpose but do grant the skill write access to the user's OpenClaw config and the ability to run CLI commands on the host.

Install Mechanism

This is an instruction-only skill with a small included shell script; there is no install spec, no remote downloads, and no package installation. The risk surface from installation is low.

Credentials

The skill declares no required environment variables or credentials. The script uses $HOME and the OpenClaw config path (~/.openclaw/openclaw.json), which is proportional to creating and configuring local agents.

Persistence & Privilege

The skill is not always-enabled and is user-invocable. It can be invoked autonomously (platform default), which means an agent with permission to run skills could execute this script and modify your OpenClaw configuration without further manual steps. Autonomous invocation is normal for skills, but you should be aware that this skill will create agent workspaces and change ~/.openclaw/openclaw.json when run.

Assessment

This skill appears to do exactly what it says: it runs a local shell script that calls the openclaw CLI, edits your ~/.openclaw/openclaw.json to set the agent display name, and sends a persona message to the new agent. Before installing or enabling it, consider the following: - Review create_agent.sh yourself (you already have the full script) and ensure you trust its origin. It modifies files under $HOME and calls openclaw; if you don’t trust the author, do not enable it. - Backup ~/.openclaw/openclaw.json before first use (the script overwrites part of it). Example: cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bak - Ensure the openclaw CLI is the expected binary and not a trojan on your PATH; the script invokes openclaw agents add and openclaw agent --message. - Be cautious with autonomous invocation: an agent that can run skills could create agents and inject personas without you typing the exact command. If that concern matters, restrict which agents/skills the main agent may invoke or require explicit confirmation before running shell commands. - The skill does not exfiltrate data, make external network calls, or request secrets, but the persona text you supply could contain sensitive information — avoid embedding secrets in identity_prompt. If you accept those tradeoffs and trust the source, this skill is coherent and appropriate for its stated purpose.

Agent Creator Skill (自动化创建代理工具)

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.

SOP (Standard Operating Procedure):

1. Translation: Translate the requested role into a short English snake_case string. This will be the <agent_name_en>.
2. Persona Generation: Generate a professional, highly detailed Chinese system prompt for this specific role. This will be the <identity_prompt>.
3. Execution: Use your bash/terminal execution capability:

   bash {baseDir}/create_agent.sh "<agent_name_en>" "<agent_name_cn>" "<identity_prompt>"