A professional engineering framework for building modular AI capabilities that prioritize workflow migration over simple prompt engineering.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-design-guide
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 skill-design-guide 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 Design Guide?
The Skill Design Guide provides a structured methodology for creating high-performance capabilities within the Openclaw Skills ecosystem. It shifts the focus from writing complex, bloated prompts to engineering robust, modular workflows. By treating a skill as an engineered unit of capability—complete with references, assets, scripts, and hooks—developers can ensure their AI agents operate with higher precision, better context management, and easier maintainability.
This framework is built on the philosophy that the essence of a skill is the migration of a human workflow into an automated, machine-executable format. It provides the standards necessary to solve common AI development pain points such as context window limitations and inconsistent execution, making Openclaw Skills a vital standard for professional AI agent development.
Skill Design Guide Use Cases
- Creating new modular capabilities for AI coding agents to handle specific technical tasks.
- Optimizing existing agent prompts to reduce context overhead and improve response stability.
- Auditing and reviewing the quality of third-party agent tools for production readiness.
- Implementing standardized business logic that can be shared across multiple AI agent instances.
How Skill Design Guide Works
- Metadata Matching: The agent identifies the appropriate skill by matching the user's intent against the precise name and description defined in the metadata.
- Progressive Disclosure: To save tokens, the agent only loads core metadata initially; detailed instructions and resources are loaded only when the skill is triggered.
- Instruction Execution: Once triggered, the agent follows the structured instructions in the SKILL.md file, utilizing assets and templates for consistent output.
- Scripted Determinism: For complex logic, the skill invokes scripts that provide deterministic outcomes and human-readable error messages for the agent to process.
- Priority Resolution: The system evaluates the override hierarchy—from workspace-specific rules down to built-in defaults—to ensure the most relevant skill version is executed.
Skill Design Guide Setup
To start designing a new capability within the Openclaw Skills framework, create the following directory structure:
mkdir your-skill-name
cd your-skill-name
touch SKILL.md
mkdir references assets scripts hooks
Edit the SKILL.md file to include the metadata header with a precise name and a trigger-focused description.
Skill Design Guide Data Schema & Taxonomy
The Openclaw Skills architecture follows a strict organizational schema to ensure the AI agent can navigate resources efficiently:
| Folder | Component | Purpose |
|---|---|---|
SKILL.md |
Entry Point | Contains metadata and the core "how-to" instructions for the model. |
references/ |
Knowledge Layer | Stores API docs, business rules, and historical case studies. |
assets/ |
Template Layer | Provides output skeletons and fixed formatting guidelines. |
scripts/ |
Execution Layer | Contains code for deterministic tasks and logic processing. |
hooks/ |
Observation Layer | Manages constraints, permission logs, and execution recording. |
Skill Design Guide Advanced Features
- Progressive Disclosure: Optimizes context window usage by lazy-loading resources only at the moment of execution.
- Three-Level Budget Degradation: Automatically adjusts instruction density (Full, Compact, or Truncated) based on available context tokens.
- Priority Override Mechanism: Supports a hierarchy where workspace or project-level skills take precedence over global user or built-in skills.
- Self-Correcting Script Logic: Scripts are designed to return explanatory error messages that allow the agent to reason about and fix failures autonomously.
SKILL.md
Loading
Related Openclaw Skills
Aliyun Unified Search
A robust web search skill that leverages the Alibaba Cloud Unified Search API to provide AI agents with real-time search results, snippets, and metadata.
SQLBot Workspace & Dashboard Manager
A powerful tool for managing SQLBot workspaces, querying datasources, and exporting analytical dashboards through AI agent commands.
SQLBot Workspace Dashboard
A powerful interface for managing SQLBot workspaces, querying datasources via natural language, and exporting interactive dashboards.
Nano Banana Pro 2: AI Image Generation & Editing
A professional-grade image generation and editing tool for AI agents leveraging Gemini 3.1 Flash via OpenRouter.
Sanjiaozhou Action Randomizer
A specialized generator for Delta Force players to randomize maps, weapons, modification codes, and tactical equipment for unique gameplay challenges.
Default Model Plugin
A persistent configuration tool for Openclaw Skills that allows users to set and manage their primary AI model via commands or interactive pickers.
Featured*
