Self-Improving Agent
一个面向 Automation 场景的 Agent 技能。原始说明:Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
一个面向 Automation 场景的 Agent 技能。原始说明:A callable skill factory for programmatic skill creation. Use when another skill needs to create a new skill, or when automating skill generation workflows....
name: create-skill
description: A callable skill factory for programmatic skill creation. Use when another skill needs to create a new skill, or when automating skill generation workflows. Accepts structured parameters (name, description, instructions, resources) and generates a complete, validated skill package. Triggers on programmatic calls from other skills or automation workflows that need to spawn new skills.
A callable skill factory designed for programmatic invocation by other skills.
This skill is typically called by other skills with structured parameters. Call it when you need to generate a new skill programmatically.
When calling this skill, provide the following parameters:
| Parameter | Required | Description |
|-----------|----------|-------------|
| name | Yes | Skill name (lowercase, hyphens only, <64 chars) |
| description | Yes | Skill description for frontmatter (include what it does + when to use) |
| instructions | Yes | Markdown body content for SKILL.md |
| scripts | No | List of script files to create: [{name, content, language}] |
| references | No | List of reference docs: [{name, content}] |
| assets | No | List of asset files: [{name, content}] |
| output_dir | No | Output directory (default: ~/.openclaw/workspace/skills) |
| validate | No | Run validation after creation (default: true) |
| package | No | Generate .skill file after creation (default: true) |
Call this skill with a structured request:
{
"name": "my-skill",
"description": "Brief description of what this skill does and when to use it.",
"instructions": "## Overview\n\nDetailed instructions in markdown...",
"scripts": [
{"name": "main.py", "content": "#!/usr/bin/env python3\n...", "language": "python"}
],
"references": [
{"name": "api-docs.md", "content": "# API Reference\n..."}
]
}
skill_name/ with appropriate subdirectoriesReturns a result object:
{
"success": true,
"skill_path": "~/.openclaw/workspace/skills/my-skill",
"skill_file": "~/.openclaw/workspace/skills/my-skill.skill",
"files_created": ["SKILL.md", "scripts/main.py"]
}
A calling skill might invoke this with:
Create a skill named "pdf-watermarker" that:
- Description: "Add watermarks to PDF files. Use when processing PDFs for branding or security."
- Instructions: Basic watermarking workflow with pdfplumber
- Scripts: A watermark.py script using PyPDF2
The skill factory handles all file creation, validation, and packaging.
create_skill.py - Main skill creation scriptvalidate_skill.py - Validation utilities (wraps skill-creator's package_skill.py)skill-spec.md - AgentSkills specification reference