夜雨聆风学习资源
Anthropic Academy ｜ Agent Skills 官方课程
https://anthropic.skilljar.com/introduction-to-agent-skills/434527
这是 Anthropic 官方 Agent Skill 培训课程的第5节。小何继续搬运中英双语资料，并补充自己的学习说明。这一节主要来学习如何将自己做的技能分享出去，为更多人所用。
Sharing skills 分享技能
What you'll learn 你将学到什么
Estimated time: 20 minutes预计时间：20 分钟
By the end of this lesson you'll be able to:完成本课程后，你将能够：
- Share skills with your team by committing them to a Git repository 通过将技能提交到 Git 仓库，与团队成员共享技能
- Distribute skills across projects through plugins and marketplaces 通过插件和市场跨项目分发技能
- Deploy skills organization-wide using enterprise managed settings 使用企业托管设置，在全组织范围内部署技能
- Configure custom subagents to use specific skills 配置自定义子代理以使用特定技能
Sharing skills 分享技能
Skills become much more valuable when they're shared across a team or organization. This part covers the three main distribution methods — repository commits, plugins, and enterprise managed settings — and explains how to configure custom subagents to use skills. You'll learn which approach fits which scenario and how to handle an important gotcha: subagents don't inherit skills automatically.当技能在团队或组织内共享时，其价值会大幅提升。介绍三种主要分发方式——仓库提交、插件和企业托管设置——并说明如何配置自定义子代理来使用技能。您将了解哪种方法适用于哪种场景，以及如何处理一个重要陷阱：子代理不会自动继承技能。
Key takeaways 关键要点
- Project skills in
.claude/skillsare shared automatically through Git — anyone who clones the repo gets them.claude/skills中的项目技能通过 Git 自动共享——克隆仓库的任何人都会获得这些技能 - Plugins let you distribute skills across repositories via marketplaces for broader community use 插件让您能够通过市场跨仓库分发技能，以便更广泛地供社区使用
- Enterprise managed settings deploy skills organization-wide with the highest priority, ideal for mandatory standards and compliance 企业级托管设置以最高优先级在全组织范围内部署技能，适用于强制性标准和合规要求
- Subagents don't automatically see your skills — you must explicitly list skills in a custom agent's frontmatter
skillsfield 子代理不会自动看到你的技能——你必须在自定义代理的前置元数据skills字段中明确列出技能 - Built-in agents (Explorer, Plan, Verify) can't access skills at all — only custom subagents defined in
.claude/agentscan 内置代理（Explorer、Plan、Verify）完全无法访问技能——只有在.claude/agents中定义的自定义子代理才能访问
Skills become much more valuable when they're shared. A PR review skill that only you use is helpful, but that same skill shared across your entire team standardizes code review and creates a consistent experience across your organization. Let's look at the different ways you can distribute skills.技能在分享时才会变得更有价值。一个只有你自己使用的 PR 审查技能固然有用，但若能在整个团队中共享，就能标准化代码审查流程，为组织带来一致的体验。接下来，让我们看看分发技能的不同方式。
Committing Skills to Your Repository将技能提交到您的仓库
The simplest sharing method is committing skills directly to your repository. Place them in .claude/skills, and anyone who clones the repo gets those skills automatically — no extra installation needed.最简单的分享方式是将技能直接提交到你的仓库中。将它们放在 .claude/skills 目录下，任何克隆该仓库的人都会自动获得这些技能——无需额外安装。
When you push updates, everyone gets them on the next pull. This approach works well for:当你推送更新时，所有人下次拉取时都会收到。这种方法适用于：
- Team coding standards 团队编码规范
- Project-specific workflows项目特定工作流
- Skills that reference your codebase structure引用你代码库结构的技能
The .claude directory contains your agents, hooks, skills, and settings — all version-controlled and shared with the team through normal Git workflows..claude 目录包含您的代理、钩子、技能和设置——所有这些都通过常规 Git 工作流进行版本控制并与团队共享。
Distributing Skills Through Plugins通过插件分发技能
Plugins are a way to extend Claude Code with custom functionality designed to be shared across teams and projects. In your plugin project, create a skills directory that follows a similar file structure to the .claude directory — each skill gets its own folder with a SKILL.md file inside.插件是一种通过自定义功能扩展 Claude Code 的方式，旨在跨团队和项目共享。在你的插件项目中，创建一个遵循与 .claude 目录类似文件结构的 skills 目录——每个技能都有自己独立的文件夹，其中包含一个 SKILL.md 文件。
After you distribute your plugin to a marketplace, other users can discover and install it into Claude Code for themselves.将你的插件分发到市场后，其他用户即可自行发现并将其安装到 Claude Code 中。
This approach is best when your skills aren't too project-specific and can be useful to community members beyond your immediate team.当你的技能不局限于特定项目，且对团队之外的社区成员也有用时，这种方法最为理想。
Enterprise Deployment Through Managed Settings通过托管设置进行企业部署
Administrators can deploy skills organization-wide through managed settings. Enterprise skills take the highest priority — they override personal, project, and plugin skills with the same name.管理员可以通过托管设置在整个组织范围内部署技能。企业技能拥有最高优先级——它们会覆盖同名的个人、项目和插件技能。
The managed settings file supports features like strictKnownMarketplaces to control where plugins can be installed from:托管设置文件支持诸如 strictKnownMarketplaces 之类的功能，用于控制插件的安装来源：
"strictKnownMarketplaces": [ {"source": "github","repo": "acme-corp/approved-plugins" }, {"source": "npm","package": "@acme-corp/compliance-plugins" }]This is the right choice for mandatory standards, security requirements, compliance workflows, and coding practices that must be consistent across the organization. The keyword here is "must."这是针对必须在整个组织中保持一致的强制性标准、安全要求、合规工作流和编码实践的正确选择。这里的关键词是"必须"。
Skills and Subagents 技能与子代理
Here's something that surprises people: subagents don't automatically see your skills. When you delegate a task to a subagent, it starts with a fresh, clean context.有一件事常让人感到意外：子代理并不会自动看到你的技能。当你将任务委托给子代理时，它会从一个全新的、干净的上下文开始。
There are important distinctions to understand:需要理解几个重要区别：
- Built-in agents (like Explorer, Plan, and Verify) can't access skills at all 内置代理（如 Explorer、Plan 和 Verify）完全无法访问技能
- Custom subagents you define can use skills, but only when you explicitly list them 你定义的自定义子代理可以使用技能，但仅在你明确列出它们时才能使用
- Skills are loaded when the subagent starts, not on demand like in the main conversation 技能在子代理启动时加载，而非像主对话中那样按需加载
To create a custom subagent with skills, add an agent markdown file in .claude/agents. You can use the /agents command in Claude Code to create one interactively:要创建带有技能的自定义子代理，请在 .claude/agents 中添加一个代理 Markdown 文件。您可以在 Claude Code 中使用 /agents 命令以交互方式创建：
The generated agent file includes a skills field that lists which skills to load. Here's what the frontmatter looks like:生成的代理文件包含一个 skills 字段，用于列出要加载的技能。以下是前置元数据的内容：
---name: frontend-security-accessibility-reviewerdescription: "Use this agent when you need to review frontend code for accessibility..."tools: Bash, Glob, Grep, Read, WebFetch, WebSearch, Skill...model: sonnetcolor: blueskills: accessibility-audit, performance-check---When you delegate to this subagent, it has both skills loaded and applies them to every review. First make sure the skills exist in your .claude/skills directory, then either create a new subagent or add the skills field to an existing agent's markdown file.当你将任务委托给这个子代理时，它会加载所有技能并将其应用于每次审查。首先确保技能存在于你的 .claude/skills 目录中，然后创建一个新的子代理，或者将 skills 字段添加到现有代理的 Markdown 文件中。
This pattern works really well when:这种模式在以下情况下效果非常好：
- You want isolated task delegation with specific expertise 您希望实现具有特定专业知识的独立任务委派
- Different subagents need different skills (frontend reviewer vs. backend reviewer) 不同的子代理需要不同的技能（前端审查员与后端审查员）
- You want to enforce standards in delegated work without relying on prompts 您希望在委派的工作中强制执行标准，而不依赖提示词
Lesson reflection 课程反思
- Which sharing method (repository, plugin, enterprise) makes the most sense for the skills you've been thinking about building? 对于您一直在考虑构建的技能，哪种共享方式（仓库、插件、企业级）最为合理？
- Do you have workflows where custom subagents with specific skills would improve consistency in delegated work? 您是否有工作流程，其中具有特定技能的定制子代理能够提高委派工作的一致性？
What's next 下一步
In the final lesson, you'll learn how to troubleshoot common skill issues — from skills that don't trigger, to priority conflicts, to runtime errors — with a practical checklist you can reference anytime.在最后一课中，你将学习如何排查常见的技能问题——从技能无法触发、优先级冲突到运行时错误——并附上一份随时可参考的实用检查清单。
Feedback 反馈
As you progress through the course, we'd love to hear how you're using skills in your work, plus any feedback you may have. Share your feedback here.随着课程的深入，我们非常希望了解您在工作中如何使用技能，以及您可能提出的任何反馈。请在此处分享您的反馈。
