create skill
智能创建OpenClaw技能,提供技能发现、复用建议和自适应文件生成。
MIT-0 · Free to use, modify, and redistribute. No attribution required.
⭐ 0 · 46 · 0 current installs · 0 all-time installs
byMr.Yang@yangjinghua0127
MIT-0
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The SKILL.md describes an advisory/creator tool for discovering and scaffolding OpenClaw skills and does not request binaries, env vars, or installs that are unrelated to that purpose.
Instruction Scope
Instructions stay within the scope of searching for similar skills, collecting metadata, and generating skill files. However the guidance is broad about 'generating files' and 'updating the skill registry' without explicit safeguards or a concrete API/command; if an agent implements these steps it may write or overwrite files—review generated outputs before registering or executing created scripts.
Install Mechanism
No install spec or code files are included (instruction-only). This is low risk because nothing is downloaded or written by the skill package itself.
Credentials
No environment variables, binaries, or credentials are required. The declared requirements align with the documented purpose.
Persistence & Privilege
The skill is not always-enabled and has no install actions, so it does not request elevated persistent presence. Still, it explicitly suggests generating files and potentially scripts (JS/SH) which, if executed by the agent, could have side effects—consider limiting write/execute permissions for the agent.
Assessment
This skill is an instruction-only helper for discovering and scaffolding OpenClaw skills and appears internally consistent. Before enabling or letting the agent run it: (1) confirm the agent has limited file-system privileges so generated files cannot overwrite important data; (2) review any SKILL.md, JS, or shell files it proposes before registering or executing them; (3) backup your skills directory if you plan to allow automatic updates to the skill registry; (4) be cautious if you grant the agent permission to execute created scripts—prefer manual review and explicit approval for execution.Like a lobster shell, security has layers — review code before you run it.
Current versionv1.0.0
Download ziplatest
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🎨 Clawdis
SKILL.md
Skill: 智能技能创建向导 (create-skill)
智能创建OpenClaw技能,提供技能发现、复用建议和自适应文件生成。
触发方式
创建技能
create-skill
新建技能
skill create
需要创建技能
能力
- 智能技能发现 - 创建前搜索类似技能,避免重复造轮子
- 复用建议系统 - 根据现有技能给出实用的复用建议
- 自适应文件生成 - 根据实际需求生成相应文件
- 简化创建流程 - 清晰的交互选项和指导
输出格式
根据技能需求生成相应的文件结构:
智能创建流程
1. 技能发现阶段
- 输入关键词搜索类似技能
- 显示发现结果和复用建议
- 选择创建策略(全新/扩展/取消)
2. 信息收集阶段
- 收集技能基本信息
- 选择实现方式
- 确认创建
3. 文件生成阶段
- 根据选择生成相应文件
- 更新技能注册表
- 提供后续步骤指导
示例对话
用户: 创建技能
助手: 🎯 开始执行智能技能创建向导...
🎨 请输入新技能的核心关键词:
用户输入: 文件处理
发现类似技能: 文件管理, 批量处理
是否创建新的技能?
收集技能信息...
生成技能文件...
✅ 技能创建完成
限制
- 仅在OpenClaw环境中运行
最佳实践
使用建议
- 先搜索后创建 - 充分利用技能发现功能,避免重复创建
- 考虑复用 - 优先考虑扩展现有技能,提高开发效率
- 明确需求 - 清晰描述技能功能,便于技能发现
- 逻辑极致精简 - 技能实现方案有限仅考虑纯skill文档指导,实在是需要脚本才考虑编写js或者shell脚本
- 遵循标准 - 使用YAML frontmatter标准,确保技能元数据完整
OpenClaw YAML Frontmatter 规范
每个OpenClaw技能都应包含标准的YAML frontmatter:
---
name: skill-name
description: "技能描述"
homepage: https://example.com (可选)
metadata: { "openclaw": { "emoji": "🎨", "requires": { "bins": ["node"] } } }
---
核心字段:
name: 技能名称(小写,用连字符连接)description: 技能描述(50字以内)homepage: 相关主页URL(可选)metadata.openclaw: OpenClaw元数据emoji: 技能表情符号(如🎨、📊、⚡等)requires: 依赖要求bins: 需要的二进制文件(如node、bash、curl)node_modules: 需要的npm包env: 需要的环境变量
示例:
- 文档指导型:
metadata: { "openclaw": { "emoji": "📚" } } - Node.js技能:
metadata: { "openclaw": { "emoji": "⚡", "requires": { "bins": ["node"] } } } - Shell技能:
metadata: { "openclaw": { "emoji": "🐚", "requires": { "bins": ["bash"] } } }
FAQ
技能创建的基本流程是什么?
输入关键词 → 搜索类似技能 → 收集技能信息 → 生成文件
如何避免重复创建已有功能?
使用智能技能发现功能,创建前会搜索类似技能并给出复用建议。
技能应该包含哪些文件?
- 必须:
SKILL.md(技能元数据文档) - 可选:
<技能名>.js(如果需要Node.js实现)或<技能名>.sh(如果需要Shell实现) - 可选:
package.json(如果需要脚本则需要项目配置)
如何测试新创建的技能?
- 查看技能文档:
cat skills/<技能名>/SKILL.md - 如有脚本,运行测试:
cd skills/<技能名> && node <技能名>.js - 在OpenClaw中尝试触发词
Files
1 totalSelect a file
Select a file to preview.
Comments
Loading comments…
