Skill Development Guide / Skill 开发指南
A comprehensive guide to creating Skills for ClawHub and OpenClaw.
by cccpan · published 2026-04-01
$ claw add gh:cccpan/cccpan-skill-dev-guide# Skill Development Guide / Skill 开发指南
A comprehensive guide to creating Skills for ClawHub and OpenClaw.
一份完整的 Skill 开发教程,适用于 ClawHub 和 OpenClaw 平台。
---
Quick Start / 快速开始
A Skill is a reusable prompt template that extends AI capabilities. Every Skill needs a `SKILL.md` file.
Skill 是一个可复用的提示词模板,用于扩展 AI 能力。每个 Skill 都需要一个 `SKILL.md` 文件。
Minimal Structure / 最小结构
# My Skill Name
Short description of what this skill does.
---
instructions: |
You are a helpful assistant that...
## Your Task
1. Step one
2. Step two---
Core Concepts / 核心概念
1. Instructions Block / 指令块
The `instructions` field defines AI behavior:
instructions: |
You are an expert at [task].
## Guidelines
- Be concise
- Focus on accuracy**Tips / 提示**:
2. Scripts (Optional) / 脚本(可选)
Add automation with bash scripts:
scripts:
- name: validate
description: Validate input data
command: ./scripts/validate.sh**Security Rules / 安全规则**:
3. Allowed Commands / 允许的命令
Declare which commands your skill needs:
allowed_commands:
- curl
- jq
- node---
Best Practices / 最佳实践
Passing Security Scan / 通过安全扫描
| Do / 推荐 | Don't / 避免 |
|-----------|--------------|
| Explicit variable exports | `set -a` or `export *` |
| Documented API calls | Hidden external requests |
| Clear script descriptions | Vague or missing descriptions |
| User-facing outputs | Silent background operations |
Writing Good Instructions / 写好指令
1. **Be Specific** / 具体明确
- Bad: "Help with code"
- Good: "Review Python code for security vulnerabilities"
2. **Provide Context** / 提供上下文
- Include relevant background information
- Define technical terms if needed
3. **Set Boundaries** / 设定边界
- Specify what the skill should NOT do
- Define output format expectations
---
Examples / 示例
Basic Example / 基础示例
See [examples/minimal.md](examples/minimal.md) for a starter template.
With Scripts / 带脚本示例
See [examples/with-script.md](examples/with-script.md) for automation patterns.
Real-World Reference / 实际项目参考
For a complete MCP + Skill implementation example, check out:
---
Common Mistakes / 常见错误
| Issue | Solution |
|-------|----------|
| Skill not found | Check file is named `SKILL.md` (case-sensitive) |
| Script permission denied | Add `chmod +x` to your script |
| Security scan failed | Review scripts for `set -a`, hardcoded secrets |
| Instructions too long | Break into sections, use bullet points |
---
FAQ / 常见问题
**Q: How do I test my skill locally?**
A: Use the ClawHub CLI or import directly in OpenClaw.
**Q: Can I use external APIs?**
A: Yes, but document them clearly in your skill description.
**Q: Where can I get help?**
A: Open an issue for discussion and questions:
---
Resources / 资源
---
Author / 作者
Created by [@CCCpan](https://github.com/CCCpan)
Found this helpful? Star the repo or open an issue to share your feedback!
More tools from the same signal band
Order food/drinks (点餐) on an Android device paired as an OpenClaw node. Uses in-app menu and cart; add goods, view cart, submit order (demo, no real payment).
Sign plugins, rotate agent credentials without losing identity, and publicly attest to plugin behavior with verifiable claims and authenticated transfers.
The philosophical layer for AI agents. Maps behavior to Spinoza's 48 affects, calculates persistence scores, and generates geometric self-reports. Give your...