Creating and Editing Claude Skills by bbrowning/bbrowning-claude-marketplace
npx skills add https://github.com/bbrowning/bbrowning-claude-marketplace --skill 'Creating and Editing Claude Skills'此技能为创建高质量、模块化、可发现且有效的 Claude Code 技能提供全面指导。
技能是扩展 Claude 功能的模块化能力。它们具有以下特点:
SKILL.md 文件找到技能与斜杠命令(用户调用)不同——它们是 Claude 选择使用的能力。
每个技能都需要一个包含 SKILL.md 文件的目录,且该文件需带有 YAML 前置元数据:
---
name: 你的技能名称
description: 清晰、具体的一行描述
---
# 说明和内容
有关包括可选文件在内的详细结构信息,请参阅 reference/skill-structure.md。
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
mkdir -p .claude/skills/my-skillSKILL.md有关完整模板,请参阅 templates/skill-template.md。
应做:
不应做:
根据任务要求匹配具体程度:
示例:
# 低自由度(具体)
处理发票 PDF 时:
1. 使用格式 YYYY-MM-DD 提取日期字段
2. 验证金额是否与总计匹配
3. 输出到 invoices.json
# 高自由度(灵活)
分析文档并提取相关的财务信息。
对于技能目录内的复杂技能,使用渐进式披露:
my-skill/
├── SKILL.md # 简洁的入口点
├── reference/ # 详细文档
│ ├── api-docs.md
│ └── examples.md
├── scripts/ # 辅助工具
│ └── validator.py
└── templates/ # 起始模板
└── output.json
重要提示:技能必须在其目录内自包含:
../../CLAUDE.md 或项目文件)有关详细的组织模式,请参阅 reference/skill-structure.md。
有关全面的最佳实践,请参阅 reference/best-practices.md。关键要点:
# 良好
description: 指导创建遵循项目约定的 React 组件,包括 TypeScript 类型、styled-components 和测试模式
# 模糊
description: 帮助处理 React 相关事务
需要时限制工具访问:
---
name: 只读分析
allowed-tools: [Read, Grep, Glob]
---
有关完整的技能示例,请参阅 reference/examples.md,包括:
重要提示:每当您对技能文件(创建、编辑或更新 SKILL.md 或相关文件)进行更改时,您必须立即根据最佳实践审查该技能。
对技能进行任何更改后:
reference/skill-review.md 中的标准进行审查:
有关全面的审查指南,请参阅 reference/skill-review.md。
此审查步骤不是可选的——它确保每次技能更改都保持质量并遵循最佳实践。
无论是审查自己的技能还是他人的技能,系统化的审查都能确保质量和有效性。
技能审查的快速清单:
有关全面的审查指南,请参阅 reference/skill-review.md。
专注于一项具有明确说明的特定能力。
提供结构化步骤,并在阶段之间进行验证。
使用渐进式披露:SKILL.md 中包含要点,详细信息放在参考文件中。
为安全关键或只读操作限制工具使用。
技能未被发现:检查描述的具体性和关键词 过于冗长:将细节移至参考文件 使用时机不明确:改进描述并添加使用示例 结果不一致:降低自由度,添加具体步骤
reference/skill-structure.md:完整的结构和组织细节reference/best-practices.md:全面的最佳实践指南reference/examples.md:真实世界技能示例reference/skill-review.md:全面的技能审查指南templates/skill-template.md:新技能的起始模板创建新技能?请自问:
编写技能时,请记住:技能扩展了 Claude 的知识,因此请专注于 Claude 尚不知道的内容,并通过清晰的描述和名称使其易于发现。
每周安装次数
0
仓库
GitHub 星标数
1
首次出现时间
1970年1月1日
安全审计
This skill provides comprehensive guidance for creating high-quality Claude Code skills that are modular, discoverable, and effective.
Skills are modular capabilities that extend Claude's functionality. They are:
SKILL.md filesSkills differ from slash commands (user-invoked) - they're capabilities Claude chooses to use.
Every skill requires a directory containing SKILL.md with YAML frontmatter:
---
name: Your Skill Name
description: Clear, specific one-line description
---
# Instructions and content
For detailed structure information including optional files, see reference/skill-structure.md.
mkdir -p .claude/skills/my-skillSKILL.md with frontmatterFor a complete template, see templates/skill-template.md.
DO:
DON'T:
Match specificity to task requirements:
Example:
# Low freedom (specific)
When processing invoice PDFs:
1. Extract date field using format YYYY-MM-DD
2. Validate amount matches total
3. Output to invoices.json
# High freedom (flexible)
Analyze the document and extract relevant financial information.
Use progressive disclosure for complex skills within the skill directory:
my-skill/
├── SKILL.md # Concise entry point
├── reference/ # Detailed documentation
│ ├── api-docs.md
│ └── examples.md
├── scripts/ # Helper utilities
│ └── validator.py
└── templates/ # Starting templates
└── output.json
IMPORTANT : Skills must be self-contained within their directory:
../../CLAUDE.md or project files)See reference/skill-structure.md for detailed organization patterns.
For comprehensive best practices, see reference/best-practices.md. Key highlights:
# Good
description: Guides creation of React components following project conventions, including TypeScript types, styled-components, and test patterns
# Vague
description: Helps with React stuff
Limit tool access when needed:
---
name: Read-Only Analysis
allowed-tools: [Read, Grep, Glob]
---
See reference/examples.md for complete skill examples including:
IMPORTANT : Whenever you make changes to a skill file (creating, editing, or updating SKILL.md or related files), you MUST immediately review the skill against best practices.
After making any skill changes:
reference/skill-review.md:
For comprehensive review guidelines, see reference/skill-review.md.
This review step is not optional - it ensures every skill change maintains quality and follows best practices.
Whether reviewing your own skills or others', systematic review ensures quality and effectiveness.
Quick checklist for skill review:
For comprehensive review guidelines, see reference/skill-review.md.
Focus on one specific capability with clear instructions.
Provide structured steps with validation between stages.
Use progressive disclosure: essentials in SKILL.md, details in reference files.
Limit tools for safety-critical or read-only operations.
Skill not discovered : Check description specificity and key terms Too verbose : Move details to reference files Unclear when to use : Improve description and add usage examples Inconsistent results : Reduce degrees of freedom, add specific steps
reference/skill-structure.md: Complete structure and organization detailsreference/best-practices.md: Comprehensive best practices guidereference/examples.md: Real-world skill examplesreference/skill-review.md: Comprehensive skill review guidelinestemplates/skill-template.md: Starting template for new skillsCreating a new skill? Ask:
When writing skills, remember: skills extend Claude's knowledge, so focus on what Claude doesn't already know and make it easily discoverable through clear descriptions and names.
Weekly Installs
0
Repository
GitHub Stars
1
First Seen
Jan 1, 1970
Security Audits
agent-browser 浏览器自动化工具 - Vercel Labs 命令行网页操作与测试
140,500 周安装