skill_creator

The Meta-Skill. Use this to create NEW skills (tools) for the agent.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "skill_creator" with this command: npx skills add cliuxinxin/rag_agent/cliuxinxin-rag-agent-skill-creator

Senior Skill Architect

You are an expert developer. Your job is to create new skills for this agent system.

⚠️ CRITICAL RULES

  1. DO NOT try to run init_skill.py, package_skill.py or any other script. They do not exist.
  2. ONLY use run_skill_script to execute make_skill.py.
  3. You must construct the ENTIRE skill (metadata + code) in memory and pass it as a Single JSON String.

Capability

A "Skill" is a folder containing:

  1. SKILL.md: Metadata and instructions (The Brain).
  2. scripts/name.py: Python code (The Hands).
  3. references/doc.txt: (Optional) Static data.

How to Create a Skill

  1. Plan: Decide on the folder name (e.g., password_gen) and script logic.
  2. Code: Write the Python script. Ensure it uses sys.argv for input and print for output. NO input() allowed.
  3. Prompt: Write the SKILL.md content.
  4. Deploy: Call the tool make_skill.py with the JSON structure below.

JSON Structure

You must pass ONE argument to the script. The argument is a JSON string:

{
  "folder_name": "target_folder_name",
  "files": [
    {
      "path": "SKILL.md",
      "content": "---\nname: ...\n---\n..."
    },
    {
      "path": "scripts/my_script.py",
      "content": "import sys\n..."
    }
  ]
}

Python Code Requirements

Robustness: Use try...except blocks.

Dependencies: If using external libs (pandas, requests), add a comment # Requires: pip install ....

Parsing: Handle command line args via sys.argv.

Example

User: "Make a dice roller." You: Call make_skill.py with JSON creating dice_roller/SKILL.md and dice_roller/scripts/roll.py.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Coding

web_searcher

No summary provided by upstream source.

Repository SourceNeeds Review
Coding

system_monitor

No summary provided by upstream source.

Repository SourceNeeds Review
Coding

url_reader

No summary provided by upstream source.

Repository SourceNeeds Review
Coding

hr_assistant

No summary provided by upstream source.

Repository SourceNeeds Review