A comprehensive framework for creating modular, self-contained skills that transform general-purpose AI into specialized agents with domain expertise.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install qclaw-skill-creator
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install qclaw-skill-creator using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Skill Creator?
The Skill Creator provides a structured methodology for developing Openclaw Skills, which act as specialized onboarding guides for AI agents. By bundling procedural knowledge, tool integrations, and domain-specific logic, these skills allow Claude to handle complex, multi-step procedures that go beyond its base training. The system emphasizes token efficiency through a progressive disclosure design, ensuring that the AI only loads the information it needs for the specific task at hand.
Developing Openclaw Skills involves more than just writing instructions; it is about creating a robust package of scripts, reference materials, and assets. This modular approach ensures that agents remain concise and effective, avoiding context bloat while providing the deterministic reliability required for professional environments.
Skill Creator Use Cases
- Extending AI capabilities with company-specific business logic or schemas.
- Creating reusable workflows for complex file manipulations like PDF or DOCX editing.
- Providing specialized API documentation and reference materials for technical tasks.
- Standardizing multi-step procedures to ensure consistency across different agent sessions.
- Reducing token usage by offloading verbose documentation into load-on-demand reference files.
How Skill Creator Works
- Discovery: Analyze concrete examples of how the skill will be used to identify necessary functionality.
- Planning: Determine which reusable resources (scripts, references, assets) are required to support the workflows.
- Initialization: Use the mandatory init_skill.py script to generate the standardized directory structure and SKILL.md template.
- Implementation: Populate the skill with executable scripts, domain-specific markdown references, and static assets.
- Optimization: Refine the SKILL.md body to focus on essential procedural guidance while keeping it under the 500-line recommended limit.
- Validation & Packaging: Run the packaging script to validate frontmatter and structure before generating a distributable .skill file.
- Iteration: Test the skill in real-world scenarios and update resources based on performance feedback.
Skill Creator Setup
To create a new skill, you must use the provided initialization script. Never create skill directories manually.
# Standard initialization for Openclaw Skills
scripts/init_skill.py <skill-name> --path ~/.qclaw/skills
# For Windows users
python scripts\init_skill.py <skill-name> --path %USERPROFILE%\.qclaw\skills
# Packaging a completed skill
scripts/package_skill.py <path/to/skill-folder>
Skill Creator Data Schema & Taxonomy
Every skill follows a strict hierarchical directory structure to ensure compatibility and efficiency.
| Component | File/Folder | Purpose |
|---|---|---|
| Metadata | SKILL.md (Frontmatter) | Contains the name and description used for skill triggering. |
| Instructions | SKILL.md (Body) | Markdown-based procedural guidance and workflow instructions. |
| Scripts | scripts/ |
Executable Python or Bash code for deterministic tasks. |
| References | references/ |
Technical documentation and schemas loaded into context on demand. |
| Assets | assets/ |
Static files, templates, or icons used in the final output. |
Skill Creator Advanced Features
- Progressive Disclosure: Uses a three-level loading system (Metadata > Body > Resources) to minimize context window consumption.
- Conditional Loading: Pattern-based organization allows the agent to read only specific domain files (e.g., loading
aws.mdonly when AWS is mentioned). - Deterministic Scripting: Offloads repetitive or fragile logic to standalone scripts that can be executed without filling the context window.
- Automated Validation: The packaging process includes a built-in validator to check for YAML syntax, naming conventions, and resource link integrity.
- Variable Freedom Levels: Design guidelines for setting High, Medium, or Low degrees of freedom based on task fragility.
SKILL.md
Loading
Related Openclaw Skills
qclaw-text-file
A high-priority text-writing skill that replaces standard write tools to ensure perfect cross-platform encoding and newline compatibility.
Tencent Survey
A professional integration for managing Tencent Survey (wj.qq.com) operations including survey creation, structural updates, and data retrieval via AI agents.
TME OpenAPI Skill
A universal client for discovering and invoking Tencent Music OpenAPI operator services with fully automated authentication.
Video Compressor 8MB
A high-performance AI video compression skill that automatically optimizes large video files to fit under strict 8MB limits using cloud GPU rendering.
Crystal Geyser
Crystal Geyser is an Openclaw Skills module providing deep technical and business insights into one of America's most successful independent bottled water brands, CG Roxane.
Bilingual Humanizer
A professional editing tool designed to identify and remove AI-generated signatures, making text sound authentically human in both Spanish and English.
Featured*
