以下为您整理的章节 1内容。

章节 1：引言与核心价值 (Introduction & Core Value)

📝 本节摘要：

本节作为开篇，主讲人 Rashid 介绍了 Anthropic 发布的新功能 Claude Skills，强调其能将原本耗时 30-60 分钟的工作自动化缩减至 5 分钟。同时给出了本期视频的三个核心议程：辨析 Skills、Projects 与普通对话的区别；演示在 Web 端创建 Skill 的流程；以及如何在 Claude Code 中使用 Skills 实现更高级的自主代理（Agent）功能。主讲人特别指出，无需编程背景即可掌握此技术，旨在帮助用户节省时间并创造更多营收。

[原文] [Rashid]: just about a month ago Anthropic released this feature called Claude Skills i've been using it for the past month and it completely transformed how I get work done in my business allowing me to automate work that would take me around 30 minutes sometimes an hour to get done in just around 5 minutes

[译文] [Rashid]: 就在大约一个月前，Anthropic 发布了一项名为 Claude Skills（Claude 技能）的功能。过去一个月我一直在使用它，它彻底改变了我在业务中完成工作的方式，让我能够将原本需要 30 分钟甚至一小时才能完成的工作，自动化缩减到仅需 5 分钟左右。

[原文] [Rashid]: so in this video I'm going to give you a complete breakdown on how to use Claude skills in your business

[译文] [Rashid]: 所以在这个视频中，我将为你全面拆解如何在你的业务中使用 Claude Skills。

[原文] [Rashid]: first I'm going to show you the difference between skills projects and just using Claude normally

[译文] [Rashid]: 首先，我会向你展示 Skills（技能）、Projects（项目）与仅仅正常使用 Claude 之间的区别。

[原文] [Rashid]: second I'll walk you through how to create a skill using the Cloud web app and I'll go over some examples on how you'd want to use Cloud Skills

[译文] [Rashid]: 其次，我会一步步带你了解如何使用 Cloud Web 应用程序创建一个 Skill，并会过一遍你想如何使用 Cloud Skills 的一些案例。

[原文] [Rashid]: and finally at the end of the video I'll show you how you could use Cloud Skills within Cloud Code so you can empower agents or use these skills with Cloud Code so you can get more work done autonomously

[译文] [Rashid]: 最后，在视频的结尾，我会向你展示如何在 Claude Code（代码环境）中使用 Cloud Skills，这样你就可以赋能智能体（Agents），或者结合 Claude Code 使用这些技能，从而自主地完成更多工作。

[原文] [Rashid]: and by the end of this video you'll learn exactly how to use Cloud Skills so you can regain back more time and use that time to generate more revenue in your business

[译文] [Rashid]: 到视频结束时，你将确切学会如何使用 Cloud Skills，这样你就能重新夺回更多时间，并利用这些时间为你的业务创造更多收入。

[原文] [Rashid]: and no you do not need to know how to code you do not need to be technical you just need to spend the next few minutes paying attention so you can learn how to set this up for yourself

[译文] [Rashid]: 还有，不，你不需要懂代码，你不需要懂技术，你只需要在接下来的几分钟里集中注意力，这样你就能学会如何为自己设置这一切。

注：原文中出现的 "Cloud Skills" 和 "Cloud web app" 结合上下文及 Anthropic 语境，实际指代 "Claude Skills" 和 "Claude web app"，译文中已做相应语义调整或保留原文发音直译。

以下为您整理的章节 2内容。

章节 2：Claude Skills 的技术解构与“剧本”隐喻 (Technical Anatomy & The 'Playbook' Metaphor)

📝 本节摘要：

本节中，Rashid 深入解析了 Claude Skills 的本质。他用“剧本”（Playbook）作比喻，解释了 Skill 实际上是一个包含指令、标准和参考资料的文件夹。技术上，核心在于 skill.md 文件，它通过 YAML 页眉（名称与描述）让 Claude 能够快速识别并按需加载，而非一次性读取所有内容。这种机制极大地提高了 Token 效率，避免了传统对话中因复制粘贴大量背景信息而导致的上下文混乱（Context Window）问题。

[原文] [Rashid]: so what is cloud skills think of cloud skills kind of like giving a playbook to someone else to get work done but instead to AI

[译文] [Rashid]: 那么什么是 Claude Skills（技能）？你可以把 Claude Skills 想象成把一本剧本（Playbook）交给别人去完成工作，只不过这次是交给 AI。

[原文] [Rashid]: now within this playbook you're probably going to have instructions maybe some references examples standards guidelines and so on

[译文] [Rashid]: 在这本剧本里，你可能会有指令，也许还有一些参考资料、示例、标准、指南等等。

[原文] [Rashid]: so cloud skills acts like a playbook but to be more technical a skill is is basically just a folder it is a folder that has a skill.md file

[译文] [Rashid]: 所以 Claude Skills 就像一本剧本，但说得更技术一点，一个 Skill 基本上就是一个文件夹，一个包含 skill.md 文件的文件夹。

[原文] [Rashid]: so a skill.md file is just a markdown file that has all the instructions and configuration on how to do this specific task or workflow

[译文] [Rashid]: skill.md 文件只是一个 Markdown 文件，其中包含了关于如何执行特定任务或工作流的所有指令和配置。

[原文] [Rashid]: now because cloud skills can act like a playbook we can treat skills kind of like our playbook library so when you ask AI to do something and there's a playbook on it that exists let's say for example you ask AI write a proposal for this client okay what AI is going to do it's going to say which playbook do I need it's going to look at your playbook library or your skills and it's going to take that exact playbook which is called proposals okay

[译文] [Rashid]: 既然 Claude Skills 可以充当剧本，我们就可以把 Skills 看作是我们的剧本库。所以当你要求 AI 做某事，且存在相关的剧本时——比方说，你让 AI“为这个客户写一份提案”，好吧，AI 会怎么做？它会问“我需要哪本剧本？”，它会查看你的剧本库或你的 Skills，然后它会提取那本确切的、名为“提案”的剧本，好吗？

[原文] [Rashid]: then from that playbook it's going to read through that playbook or the skill file and it's going to understand this is how we need to do this task and then it's going to do the task for you

[译文] [Rashid]: 然后从那本剧本中，它会通读那本剧本或 Skill 文件，它会理解“这是我们需要如何完成这项任务的方式”，然后它会为你完成这项任务。

[原文] [Rashid]: so this is awesome because instead of needing to copy paste prompts and asking AI to do things step by step you can have predefined processes or skills in these what I like to call playbooks that they could pull out from your library and do multiple tasks whenever required

[译文] [Rashid]: 这太棒了，因为你不需要复制粘贴提示词（Prompts）并要求 AI 一步步做事，你可以在这些我称之为“剧本”的东西里预定义流程或技能，它们可以随时从你的库中提取出来并执行多项任务。

[原文] [Rashid]: and you can have as many playbooks or skills as you want you can have 10 20 30 plus playbooks all listed out in your in your library and then AI could reference all of these skills whenever required

[译文] [Rashid]: 你想要多少剧本或 Skill 都可以，你可以有 10 个、20 个、30 个以上的剧本全部列在你的库中，然后 AI 可以在需要时引用所有这些 Skill。

[原文] [Rashid]: and the best thing about it is that it will only load what it needs so if you ask it just write the proposal it will only reference a proposal playbook it will not reference everything and it will lose all of its context where you need to start a new chat it will only reference what's in that folder to get that specific work done that means it's very token efficient

[译文] [Rashid]: 最棒的一点是，它只会加载它需要的东西。所以如果你要求它只是写提案，它将只引用提案剧本，它不会引用所有东西导致丢失所有上下文（Context）从而让你不得不开始新对话。它只会引用那个文件夹里的内容来完成特定的工作，这意味着它非常节省 Token（代币/计算资源）。

[原文] [Rashid]: now to be more specific how the skill anatomy looks like or how the playbook looks like is that it's basically a folder and within the folder we have a skill.md file which is a markdown file and within the skill.md file this is going to be the prompt and the instructions

[译文] [Rashid]: 现在具体来说，Skill 的解剖结构或剧本的样子，基本上就是一个文件夹，在文件夹里我们有一个 skill.md 文件，这是一个 Markdown 文件，而在 skill.md 文件里，就是提示词和指令。

[原文] [Rashid]: now we can also have some more complex folders within our skill folder which are optional that will allow AI to execute more complex workflows let's say you need to run code you can have a scripts folder let's say you have templates that you want AI to reference you can have a templates folder or if you have other references that you want AI to look at to reference you can have folders and you can have as many folders as you want whatever is required to execute that skill you can have that in there

[译文] [Rashid]: 我们还可以在 Skill 文件夹中包含一些更复杂的文件夹，这是可选的，这将允许 AI 执行更复杂的工作流。比方说你需要运行代码，你可以有一个 scripts（脚本）文件夹；比方说你有希望 AI 引用的模板，你可以有一个 templates（模板）文件夹；或者如果你有希望 AI 查看引用的其他参考资料，你可以有文件夹。你可以拥有任意数量的文件夹，只要是执行该 Skill 所需的，你都可以放在那里。

[原文] [Rashid]: but the most important thing here is a skill.md file it cannot be a skill unless we have a skill.md file that Claude can read through

[译文] [Rashid]: 但这里最重要的是 skill.md 文件。除非我们有一个 Claude 可以通读的 skill.md 文件，否则它就不能被称为一个 Skill。

[原文] [Rashid]: so if you're just getting started I recommend starting simple with just a simple skill that just has a simple prompt that you could put into a skill.md file and then test it out and then you could test out some other complex skills that will have scripts references and templates and so on

[译文] [Rashid]: 所以如果你刚开始，我建议从简入手，做一个简单的 Skill，只需要把一个简单的提示词放进 skill.md 文件里，然后测试它，之后你可以测试一些包含脚本、参考资料和模板等的更复杂的 Skill。

[原文] [Rashid]: so within the skill.mmd file it's how claude identifies whether this skill is relevant to the current task because when you have all these skills or playbooks in your library AI is going to quickly read through all of them it's not going to read through all the context within the skill it's going to read through the name and the description which is the skill.md header okay

[译文] [Rashid]: 在 skill.md 文件中，Claude 就是通过它来识别该 Skill 是否与当前任务相关。因为当你的库里有所有这些 Skill 或剧本时，AI 会快速浏览所有它们。它不会通读 Skill 里的所有上下文内容，它只会读取名称和描述，也就是 skill.md 的页眉（Header），好吗？

[原文] [Rashid]: so it's read through the name and the description now what they call this technically what this part is called is called the YAML front matter technically what it's called but you we don't need to worry about that you just need to understand that just like how you have a playbook let's say a librarian you're going to look for a specific book to do a specific task what are you going to look for you can look for the name of the book and what the book does

[译文] [Rashid]: 所以它通读名称和描述。现在，他们技术上称这部分为 YAML Front Matter（YAML 前置数据），这是技术术语，但我们不需要担心这个。你只需要理解，就像你有一本剧本，或者说作为图书管理员你要找一本特定的书来做特定的任务，你会找什么？你会找书名和这本书是干什么的。

[原文] [Rashid]: same thing with your AI agent it's going to interact with your skills by quickly reading the name and description so it knows which skill to use so that's what Claude sees first

[译文] [Rashid]: 你的 AI 智能体也是一样，它通过快速读取名称和描述来与你的 Skill 交互，这样它就知道该使用哪个 Skill。这就是 Claude 首先看到的东西。

[原文] [Rashid]: now once Claude identifies that this skill is required then it's going to read through the instructions and the prompt which is stored right below okay so this is again a markdown file i'm going to go over how this looks like in a moment once I explain completely how skills work

[译文] [Rashid]: 一旦 Claude 确定需要这个 Skill，它就会通读存储在下方的指令和提示词，好吗？这依然是一个 Markdown 文件，等我完全解释完 Skill 的工作原理后，我会马上展示它的样子。

[原文] [Rashid]: so in simple terms a skill is just a prompt with a header or metadata and again the best thing about the skills is that there's automatic routing so you do not need to tell Claude to do a specific skill you could ask it if you want to be more specific but Claude is smart enough where it could read through the skills very quickly and then identify this is the skills that's required

[译文] [Rashid]: 简而言之，Skill 就是一个带有页眉或元数据（Metadata）的提示词。再说一次，Skill 最棒的地方在于有自动路由（Automatic Routing）。所以你不需要告诉 Claude 去执行特定的 Skill——如果你想更具体一点也可以要求它——但 Claude 足够聪明，它可以非常快速地通读 Skill，然后识别出这是所需的 Skill。

[原文] [Rashid]: so for example let's say writes an email about our product launch ai will quickly understand okay do I have a skill on that it will see okay there's an email skill it'll go through the email skill it will use the email skill and then within the email skill you might have a folder that's called product launch product launch templates or example so this is great because now you might have you might even have a playbook here which is called product launch and then it's able to do a product launch for you

[译文] [Rashid]: 举个例子，比方说“写一封关于我们产品发布的邮件”。AI 会迅速理解“好的，我有关于这个的 Skill 吗？”它会看到“好的，有一个邮件 Skill”，它会浏览邮件 Skill，它会使用邮件 Skill。然后在邮件 Skill 内部，你可能有一个文件夹叫“产品发布”，里面有产品发布模板或示例。这太棒了，因为现在你甚至可能在这里就有一个叫做“产品发布”的剧本，然后它就能为你进行产品发布。

[原文] [Rashid]: but again this is very important the naming of the skills is very important because you don't want the CL cloud to be confused so if you have email then you probably want to have it an email skill and then with an email skill you can have a product launch uh workflow within it so technically you can have skills within skills it does work because AI could read through it

[译文] [Rashid]: 但再次强调，这一点非常重要：Skill 的命名非常重要，因为你不想让 Claude 感到困惑。所以如果你有邮件任务，那你可能想要一个“邮件 Skill”，然后在邮件 Skill 内部，你可以包含一个“产品发布”的工作流。所以技术上讲，你可以有“Skill 套 Skill”（Skills within Skills），这是行得通的，因为 AI 可以通读它。

[原文] [Rashid]: so and the way skills load is what makes it efficient so you simply ask Claude in a chat to do something and if there's a skill for it Claude scans the skills it picks the right skills it loads the instructions and then it executes so it's very efficient it only loads what it needs and the context stays clean

[译文] [Rashid]: Skill 加载的方式正是其高效的原因。你只需在聊天中要求 Claude 做某事，如果有对应的 Skill，Claude 会扫描这些 Skill，挑选正确的 Skill，加载指令，然后执行。所以它非常高效，它只加载它需要的内容，且上下文保持整洁。

[原文] [Rashid]: this is the best thing about skills because the old way if you wanted to do this is that let's say you have a skill that has a lot of different parts to it let's you want to write an email what you do is you probably copy paste the brand guidelines you copy paste email examples you copy paste format rules you copy paste instructions and if you put all of this into a chat or a context window you're going to run out of messages very quickly and then you're going to need to copy paste everything into a new chat and then it's gets very difficult to work

[译文] [Rashid]: 这是 Skill 最棒的地方。因为如果你想用旧方法做这件事——比方说你有一个包含很多不同部分的技能，比如你想写一封邮件——你会怎么做？你可能会复制粘贴品牌指南，复制粘贴邮件示例，复制粘贴格式规则，复制粘贴指令。如果你把所有这些都放进一个聊天或上下文窗口中，你会很快耗尽消息额度，然后你需要把所有东西复制粘贴到一个新聊天中，工作就会变得非常困难。

[原文] [Rashid]: now when we use skills what happens is you get all this free output so Claude reads a skill and doesn't take any of the tokens or the context or the messages of that chat so now it's way more efficient so you can really get work done with skills

[译文] [Rashid]: 现在当我们使用 Skills 时，发生的情况是你获得了所有这些免费的输出能力。Claude 读取一个 Skill，但这不会占用该聊天的任何 Token、上下文或消息额度。所以现在它高效得多，你可以真正利用 Skills 完成工作。

以下为您整理的章节 3内容。

章节 3：决策指南：Skills、Projects 与普通对话的边界 (Strategic Framework: When to Use Skills)

📝 本节摘要：

本节提供了判断是否需要创建 Skill 的核心决策框架。Rashid 提出了两个关键问题：任务是否重复？是否需要遵循特定流程？只有同时满足这两个条件才建议制作 Skill。他通过正反案例（如将邮件转为社媒帖适合做 Skill，而一般性的头脑风暴则不适合）划清了界限。此外，他还通过类比清晰区分了三者定位：Skills 是“受过训练的专才”，Projects 是“共享知识库”（如品牌指南），而普通对话则是“聪明的通才”用于探索性思考。

[原文] [Rashid]: so next question is what should be a skill so there's two questions if you answer both of these questions yes then you should 100% turn it into skill

[译文] [Rashid]: 那么下一个问题是，什么应该成为一个 Skill？这里有两个问题，如果你对这两个问题的回答都是“是”，那么你 100% 应该把它变成一个 Skill。

[原文] [Rashid]: question number one do you do this task repeatedly and do you want clot to follow your specific process if you answered yes to both these questions then yes you should make it a skill

[译文] [Rashid]: 问题一，你是否重复执行这项任务？（问题二）你是否希望 Claude 遵循你的特定流程？如果你对这两个问题都回答“是”，那么是的，你应该把它做成一个 Skill。

[原文] [Rashid]: now the best thing about skill is that anything can be turned into skill as long as it's it's a step-by-step process that you want the AI to follow it can be a skill

[译文] [Rashid]: 关于 Skill 最棒的一点是，任何东西都可以变成 Skill。只要它是一个你希望 AI 遵循的逐步流程，它就可以是一个 Skill。

[原文] [Rashid]: so let's say you want to turn email into social posts that can be a skill if you want to have an intake form and turn that into proposal that could be a skill if you want to turn meeting notes into action items that could be a skill if you want to turn specific content that AI had written and you want to apply your brand voice to that you can also have a skill for that anything that has a process a specific style or way that you do it can be turned into a skill

[译文] [Rashid]: 举个例子，假设你想把邮件变成社交媒体帖子，这可以是一个 Skill；如果你想把一份需求表单（Intake Form）变成一份提案，这可以是一个 Skill；如果你想把会议记录变成行动项，这可以是一个 Skill；如果你想把 AI 写好的特定内容应用上你的品牌声音，你也可以为此做一个 Skill。任何有流程、特定风格或你特定的执行方式的事情，都可以变成一个 Skill。

[原文] [Rashid]: now what should you not use for skills let's say you want to ask AI what to prioritize probably don't need a skill for that unless you have a prioritization style or process that you do think then you'd have a skill for that

[译文] [Rashid]: 那么什么不该用 Skill 呢？比方说你想问 AI 该优先处理什么，这可能不需要 Skill，除非你有一套特定的优先级排序风格或流程，那样你才需要为此建一个 Skill。

[原文] [Rashid]: review my strategy you want to ask to review your strategy probably won't need a skill for that unless you have a strategy review process then you'd have a skill

[译文] [Rashid]: “审查我的策略”，你想让它审查你的策略，这可能不需要 Skill，除非你有一套策略审查流程，那你才会有个 Skill。

[原文] [Rashid]: problem solving if you want a novel problem solving probably won't need a skill unless you have like a problem solving framework uh research probably won't need a skill for that unless you have like a technique or a way that you do research then you turn into skill okay

[译文] [Rashid]: 解决问题，如果你想要一种新颖的问题解决方案，可能不需要 Skill，除非你有一套解决问题的框架；调研，这可能不需要 Skill，除非你有某种特定的调研技巧或方式，那你才把它变成 Skill，好吗？

[原文] [Rashid]: so generally if the task is repeatable and you want cloud to follow a specific process framework or style then you'd have a skill for it

[译文] [Rashid]: 所以总体而言，如果任务是可重复的，且你希望 Claude 遵循特定的流程、框架或风格，那么你就该为此设一个 Skill。

[原文] [Rashid]: so there's a lot of confusion between what how to use skills projects and just using clouds so to keep things simple we want to have skills when we we want to have AI to act like a trained specialist with a specific process framework or style simple test if you do this task repeatedly and you want AI to follow a specific process skill

[译文] [Rashid]: 现在对于如何使用 Skills、Projects（项目）以及仅仅使用 Claude 之间存在很多困惑。为了简单起见，当我们希望 AI 像一个受过训练的专家（Trained Specialist）那样，带着特定的流程、框架或风格行事时，我们使用 Skills。简单的测试：如果你重复做这个任务且希望 AI 遵循特定流程，那就是 Skill。

[原文] [Rashid]: now with projects projects a little bit different because projects is just going to be a shared knowledge base so this is going to act like reference context within multiple chats uh within multiple chat windows so this could be brand guidelines client database company strategy this isn't necessarily a skill this isn't a process this is shared reference material so simple test is this info needed for multiple different tasks

[译文] [Rashid]: 至于 Projects（项目），Projects 有点不同，因为 Projects 只是一个共享知识库。它将在多个聊天、多个聊天窗口中充当参考上下文。所以这可以是品牌指南、客户数据库、公司战略。这不一定是 Skill，这不是流程，这是共享参考资料。所以简单的测试是：这信息是否被多个不同的任务所需要？

[原文] [Rashid]: now if you just want to use cloud regularly then probably want to use it as a smart generalist or exploratory thinking then you just ask it simple questions they get back and forth judgment then you could use cloud regularly but generally at any point if you have a proprietary way of doing things a framework a process you'd want to use skills

[译文] [Rashid]: 那么如果你只是想正常使用 Claude，你可能想把它当作一个聪明的通才（Smart Generalist）或者用于探索性思考，那你只管问它简单的问题，进行来回的判断，这时你可以正常使用 Claude。但总体来说，在任何时候，如果你有一套独有的做事方式、框架或流程，你会想使用 Skills。

以下为您整理的章节 4内容。

章节 4：Web 端实操：构建“品牌声音” Skill (Web Walkthrough: Building the 'Brand Voice' Skill)

📝 本节摘要：

本节进入实操演示。Rashid 展示了如何在 Claude Web 端开启“Skill Creator”功能。他引导 AI 分析自己的工作区，并决定创建一个“品牌声音 DNA”技能。为此，他利用 Claude 的 Agent 能力连接 Notion，抓取自己过去的 10 封邮件作为训练素材。最终，Claude 自动生成了一个包含 skill.md（含 YAML 页眉）和风格分析文件的完整 Skill 文件夹，旨在让未来的 AI 生成内容都能模仿他的个人语气。

[原文] [Rashid]: so let's open up Cloud i'll give you a walkthrough on how to create your first skill

[译文] [Rashid]: 那么让我们打开 Claude，我会一步步带你了解如何创建你的第一个 Skill。

[原文] [Rashid]: once you open up Claude the first thing you want to do to create skills is you need to open up the settings and then when you go to settings you want to go to capabilities right over here and then capabilities if you scroll down you're going to see the skills menu right over here so you want to scroll down as you can see I have some skills that I've created you want to scroll down and you want to make sure you activate skill creator because what this is going to do this is going to have claude access to the skill of creating skills so um basically this is going to create the skills or playbooks for you

[译文] [Rashid]: 打开 Claude 后，创建 Skill 的第一件事是打开“设置”（Settings），然后进入“能力”（Capabilities）。向下滚动，你会看到 Skills 菜单。正如你所见，我已经创建了一些 Skill。你要向下滚动并确保激活“Skill Creator”（技能创建器），因为这会让 Claude 获得创建 Skill 的能力，基本上它会为你创建 Skill 或剧本。

[原文] [Rashid]: okay so once you have this on the next step what you could do is click on the three dots here click on try in chat so hey Claude I just added the skill creator skill can you make something amazing with it let's test that out

[译文] [Rashid]: 好的，一旦开启这个功能，下一步你可以点击这里的三圆点，点击“在聊天中尝试”（Try in chat），然后说：“嘿 Claude，我刚添加了 Skill Creator 技能，你能用它做点惊人的东西吗？”让我们测试一下。

[原文] [Rashid]: so now as you can see the thinking here it's looking through my workspace it's giving me some ideas on what skills I could create... brand voice DNA extractor i do like that analyze your existing content and extract specific patterns and make your voice unique... okay let's go with the brand voice let's create a skill for our brand voice DNA

[译文] [Rashid]: 现在你可以看到它正在思考，它正在浏览我的工作区，给我一些关于我可以创建什么 Skill 的想法……“品牌声音 DNA 提取器”，我喜欢这个，分析你现有的内容并提取特定模式，让你的声音独一无二……好的，我们就选“品牌声音”，让我们为我们的品牌声音 DNA 创建一个 Skill。

[原文] [Rashid]: before you go ahead can you let me know what source material or content you're going to use to create this... u what I'm going to do here is I'm going to ask it to go through my notion because I have this connection... so I'm going to ask it to do that let's see what it does

[译文] [Rashid]: 在你开始之前，能不能让我知道你会用什么源素材或内容来创建这个？……我要做的是，我会让它去浏览我的 Notion，因为我有这个连接……所以我让它这样做，看看它会做什么。

[原文] [Rashid]: so now claude is going through my notion and then um let's see if it finds those emails okay so now it seems like it's it's found my emails my past 10 emails perfect

[译文] [Rashid]: 现在 Claude 正在浏览我的 Notion，让我们看看它是否找到了那些邮件。好的，看来它找到了我的邮件，我过去写的 10 封邮件，完美。

[原文] [Rashid]: so as you can see here um Claude is acting as my agent so I'm giving it instructions and it's able to take action on my behalf it is able to reason understand and take action out of it so this is an example of what an agent is... AI agent basically is when you ask AI to do something and it's able to understand your instructions reason and create a plan and then execute

[译文] [Rashid]: 就像你在这里看到的，Claude 正在充当我的智能体（Agent）。我给它指令，它能够代表我采取行动，它能够推理、理解并据此采取行动。这就是智能体的一个例子……AI 智能体基本上就是当你要求 AI 做某事时，它能够理解你的指令，进行推理，制定计划，然后执行。

[原文] [Rashid]: imagine once I get the skill done every single piece of content that AI writes I could just hey can you apply uh this skill or like the brand my brand voice skill to it and then it will go through the full skill and then it will look at the text that I that AI written and it will be able to transform it into something that sounds more like me

[译文] [Rashid]: 想象一下，一旦我完成了这个 Skill，每一篇 AI 写的内容，我都可以说：“嘿，你能应用这个 Skill 或者我的品牌声音 Skill 吗？”然后它会过一遍完整的 Skill，查看 AI 写的文本，并能够将其转化为听起来更像我的内容。

[原文] [Rashid]: done here's what I built so if you look at the skill folder... it's called uh brand voice DNA there's a skill.md file which is the main skill with uh YAML header plus six layer analysis framework and then it has references and then it has my um my DNA uh in the markdown file

[译文] [Rashid]: 完成了，这是我构建的东西。如果你看这个 Skill 文件夹……它叫“品牌声音 DNA”，里面有一个 skill.md 文件，这是包含 YAML 页眉和六层分析框架的主 Skill 文件，然后它有参考资料，还有 Markdown 文件里的我的“DNA”。

[原文] [Rashid]: so this is the skill.md file it's properly structured with uh a name and then a description double check it here yeah name and then description there we go and then the next file is my brand voice DNA so the how my brand how I my content is written with instructions

[译文] [Rashid]: 这就是 skill.md 文件，它结构正确，有一个名称和一个描述。在这里复查一下，是的，名称，然后是描述，没问题。下一个文件是我的“品牌声音 DNA”，也就是关于我的品牌、我的内容是如何编写的指令。

注：此部分演示了 Claude 的 Agent 能力（主动连接 Notion 读取数据）以及 skill.md 文件的自动生成过程，为下一章的部署奠定了基础。

以下为您整理的章节 5内容。

章节 5：部署与验证：Skill 的上传与效果对比 (Deployment & Validation: Uploading & Testing)

📝 本节摘要：

Skill 创建完成后，Rashid 演示了将其投入使用的全过程：首先将生成的文件夹打包为 ZIP 压缩包，通过设置菜单上传至 Claude。为了直观展示效果，他先利用 Gemini 生成了一封平庸的“AI 垃圾内容”（AI slop）邮件，随后要求 Claude 调用刚上传的“品牌声音” Skill 对其进行重写。结果显示，Claude 通过“渐进式披露”（Progressive Disclosure）读取手册并应用风格，将原本机械的文字转化为了极具个人特色的高质量邮件。Rashid 总结道，这就像是用特定的职位描述和员工手册雇佣了一名数字员工，而不是把 AI 仅仅当作搜索引擎。

[原文] [Rashid]: so here here's what we're going to do you're going to download all yes please so we need to package it into a zip file and then from the zip file we'll be able to upload it if that doesn't work cuz previously it prompted me to add the skill and it worked but this time doesn't work so this is how you do it

[译文] [Rashid]: 所以我们现在要做的是，你要把这些都下载下来，是的请下载。我们需要把它打包成一个 ZIP 文件，然后通过 ZIP 文件我们将能够上传它。如果（自动添加）不起作用的话——因为之前它提示我添加 Skill 并且成功了，但这次没成功——所以这就是你要手动做的方法。

[原文] [Rashid]: so um open here I could do is so next thing I need to do is I need to open up settings we go to capabilities and then we go to upload skill and then I could drag and drop let's see how it works so I have this folder open let's drag it okay so it needs to be in a in a zip so I go to my downloads zip file drag it in and let's see hopefully this works boom so it's uploaded so now I could use this whenever

[译文] [Rashid]: 所以打开这里，我下一步要做的是打开“设置”，我们要去“能力”（Capabilities），然后去“上传 Skill”（Upload Skill），然后我可以拖放。让我们看看怎么操作，我打开了这个文件夹，拖进去，好的它需要是 ZIP 格式。所以我去下载文件夹，把 ZIP 文件拖进去，看看，希望这能行……Boom（嘭），上传成功了，所以现在我随时可以使用它了。

[原文] [Rashid]: so let's say try in chat and I'm going to do this i'm going to open up a let's let's do this so I'm going to do is going to open up Gemini i'm going to ask you to write can you write an email on the importance of learning how to use Claude skills in your business and how it can save 20 hours of your time so I'm going to take whatever Gemini wrote and then paste it and then test out the skill again like we're doing this live it might turn out to be like horrible but let's see

[译文] [Rashid]: 让我们点击“在聊天中尝试”。我要做的是，我要打开 Gemini，我会要求它写一封关于“在业务中学习使用 Claude Skills 的重要性以及它如何能为你节省 20 小时”的邮件。然后我会把 Gemini 写的内容拿过来，粘贴进去，再测试一下这个 Skill。我们在做直播演示，结果可能会很糟糕，但让我们拭目以待。

[原文] [Rashid]: okay you can see how this email is horrible it's literally like AI slop so let's see what what the what it does boom let's see how it works

[译文] [Rashid]: 好的，你可以看到这封邮件有多糟糕，它简直就是“AI 垃圾内容”（AI Slop）。所以让我们看看它（Claude Skill）能做什么，Boom，让我们看看它是如何工作的。

[原文] [Rashid]: okay okay so what's going on here now it's reading through that skill okay so it's reading through the skill and it's going to apply the playbook to this email let's see how it goes okay so now it's reading my voice DNA and email examples for full calibration

[译文] [Rashid]: 好的，现在发生了什么？它正在通读那个 Skill，它正在通读 Skill 并且要把这个“剧本”应用到这封邮件上。让我们看看进展如何。好的，现在它正在读取我的“声音 DNA”和邮件示例以进行全面校准。

[原文] [Rashid]: interesting i'm curious to see how this comes out so what's going on here it's doing progressive disclosure so it's not loading everything at once it's slowly progressing and seeing the context as it needs um this is great because instead of like wasting all the tokens it's very smart about the tokens

[译文] [Rashid]: 有意思，我很好奇结果会怎样。这里发生的是它正在进行“渐进式披露”（Progressive Disclosure），所以它不是一次性加载所有内容，它是缓慢推进并根据需要查看上下文。这太棒了，因为它不像那样浪费所有 Token，它在使用 Token 方面非常聪明。

[原文] [Rashid]: okay this isn't bad so it gave me subject options you're wasting your $20 a month subcl like a chat box build once runs run indefinitely you're paying $20 a month for cloud how many hours has it actually saved you if you're just opening a new chat typing a prompt and hoping for the best you're capping yourself you're treating a digital employee like a search engine

[译文] [Rashid]: 好的，这还不错。它给了我几个主题选项：“你正在浪费你每月 20 美元的订阅费”、“像聊天机器人一样只建一次然后无限运行”。（正文）：你每个月为 Claude 支付 20 美元，但它实际上为你节省了多少小时？如果你只是打开一个新聊天，输入一个提示词然后祈祷最好的结果，你就是在限制你自己。你把一个数字员工当作了搜索引擎来对待。

[原文] [Rashid]: the business owners who dominate late 2025 aren't prompting AI they're building Cloud skills what's a skill think of a standard chat as a casual conversation think of skill as hiring an employee with a specific job description and a handbook you tell Claude who to be how to act what tools to use once built you never explain the context again you just trigger it that's leverage

[译文] [Rashid]: （续正文）那些将在 2025 年底占据主导地位的企业主不是在向 AI 提问（Prompting），而是在构建 Claude Skills。什么是 Skill？把标准聊天想象成随意的交谈，把 Skill 想象成雇佣了一名带有特定职位描述和员工手册的员工。你告诉 Claude 它是谁、该如何行动、使用什么工具。一旦建成，你永远不需要再次解释上下文，你只需要触发它。这就是杠杆（Leverage）。

[原文] [Rashid]: here's where it gets interesting most people are spending two hours every week expending context to AI remember I like short sentences don't forget to format the date as blah blah blah make it sound like me every time you do that you're leaking time skills preserve context they turn off they turn one-off task into oneclick workflows i built three skills last week here's what happened see the pattern

[译文] [Rashid]: （续正文）这里变得有趣了。大多数人每周花两个小时向 AI 解释上下文：“记得我喜欢短句”、“别忘了把日期格式化为巴拉巴拉”、“让它听起来像我”。每次你这样做，你都在“泄漏”时间。Skills 保存了上下文，它们将一次性任务变成了“一键式”工作流。我上周建了三个 Skill，结果如下……看到模式了吗？

[原文] [Rashid]: wow this is really actually This is really good wow holy boly this is really good i would actually use this so this is an example of a skill brand voice

[译文] [Rashid]: 哇，这实际上真的……这真的很好，哇，天啊（Holy moly），这真的很好。我实际上真的会用这个。这就是一个 Skill 的例子：品牌声音。

以下为您整理的章节 6内容。

章节 6：进阶工作流：Claude Code 与本地环境集成 (Advanced Workflow: Integrating with Claude Code)

📝 本节摘要：

本节进入高阶应用领域。Rashid 展示了如何在 VS Code 中运行 Claude Code，实现本地文件与 AI 的深度集成。相比 Web 端，本地环境的优势在于可以手动编辑 Skill 文件，并且能够实现 Skill 之间的“链式反应”（Daisy-chaining）——即一个任务自动触发多个 Skill（如写邮件 -> 调用品牌声音 -> 存入 Notion -> 生成幻灯片）。他现场演示了如何将刚创建的“品牌声音” Skill 导入本地，并指挥 AI 自动完成从撰写邮件到生成视觉素材的全套工作流。

[原文] [Rashid]: that's how skills work in the web app let's dive into cloud code and how this would work because not only you could use this on the web app you can also use in cloud code which is even more powerful so let's dive into that

[译文] [Rashid]: 这就是 Skills 在 Web 应用中的工作方式。现在让我们深入了解 Claude Code 以及它是如何工作的，因为你不仅可以在 Web 应用上使用它，还可以在 Claude Code 中使用，这甚至更强大。让我们深入探讨一下。

[原文] [Rashid]: we're all right so this going to be the most technical part of the video so if this feels a little bit too complicated don't worry about it i'm going to be recording other videos in the future explaining how I use Cloud Code but basically to explain I'm using anti-gravity and anti-gravity is basically this application that I could have cloud code um access to my files so I have files on my computer and all of my skills the skill folders are all in my computer

[译文] [Rashid]: 好的，这部分将是视频中最技术性的部分。如果这感觉有点太复杂，别担心，我未来会录制其他视频解释我是如何使用 Claude Code 的。但简单解释一下，我正在使用 Anti-gravity（一种工具/应用），它基本上是一个让 Claude Code 能够访问我文件的应用程序。所以我的电脑里有文件，我所有的 Skills，那些 Skill 文件夹都在我的电脑里。

[原文] [Rashid]: so for example if I open up my cloud file you see agents core documentation all these folders i'm not going to go into it but I have this folder called skills okay and within these skills I have all of these skills that I have prepared previously

[译文] [Rashid]: 举个例子，如果我打开我的 Claude 文件夹，你会看到 Agents、核心文档等所有这些文件夹。我不打算深入讲那些，但我有一个叫 skills 的文件夹，好吗？在这个 skills 文件夹里，我有所有我之前准备好的 Skill。

[原文] [Rashid]: so for example if I pull up the YouTube strategy I have the skill MD file let me show you how this looks like in a uh closer look so you can see here we have the name description and you can see the description just has like many key words so um whenever I use any of these keywords it could catch that idea and then I have workflows for this skill

[译文] [Rashid]: 举个例子，如果我调出“YouTube 策略”，我有 skill.md 文件。让我给你近距离展示一下它的样子。你可以看到这里有名称、描述，你可以看到描述里有很多关键词。所以每当我使用这些关键词中的任何一个，它就能捕捉到那个想法，然后我有针对这个 Skill 的工作流。

[原文] [Rashid]: so what I'm going to do is I'm going to take this folder and I'm going to put it into here and I'm going to tell it right now so what I did was I extracted the zip folder and I moved it into my folder so if you can see it has Rashid/voice I just implemented so I'm going to ask it i just imported Rashid voice it's a skill for writing content in my voice can we implement it and integrate it into our skills withincloud folder please and can we implement routing to our other skills okay

[译文] [Rashid]: 我现在要做的是，我要拿这个文件夹（指之前下载的 ZIP），把它放进这里。我已经解压了 ZIP 文件夹并把它移到了我的文件夹里，正如你所见，这里有 Rashid/voice，我刚刚放进去的。所以我现在要告诉它：“我刚刚导入了 Rashid Voice，这是一个用我的声音写内容的 Skill。我们能不能把它实现并集成到我们 withincloud 文件夹里的 Skills 中？并且能不能实现到其他 Skill 的路由？”

[原文] [Rashid]: now the reason why I prefer cloud code is because cloud code is when I can have agents I can have skills and I can have everything in one place where these skills can daisy chain and connect to each other so for example let's say I want to write an email on my email skill i can have a step of writing an email routed to my my skill which is Rashid voice or if I write a YouTube script I can have that skill routed to this playbook so now everything is routed together kind of like playbooks

[译文] [Rashid]: 我之所以更喜欢 Claude Code，是因为在 Claude Code 里我可以拥有智能体（Agents），我可以拥有 Skills，我可以把所有东西放在一个地方，让这些 Skill 可以像菊花链（Daisy-chain）一样相互连接。举个例子，假设我想用我的邮件 Skill 写一封邮件，我可以让写邮件的步骤路由到我的 Rashid Voice Skill；或者如果我写一个 YouTube 脚本，我可以让那个 Skill 路由到这个剧本。所以现在所有东西都路由在一起，有点像剧本集。

[原文] [Rashid]: okay so Cloud Code just linked and routed all of my skills to my voice skill so I'm going to test it out i'd like to write an email on the importance of using digital assets and developing digital assets within your business... i'd like you to write this email using Yeah just write the email let's see what happens

[译文] [Rashid]: 好的，Claude Code 刚刚把我所有的 Skill 链接并路由到了我的 Voice Skill。我要测试一下。我想写一封关于“在业务中使用数字资产和开发数字资产的重要性”的邮件……我想让你写这封邮件——对，就写这封邮件，看看会发生什么。

[原文] [Rashid]: so what should happen here is that it should go through the email skill which is what I have open here and if you look at the email skill right here in the skill I have uh read Rashid voice skill so it's routed to this so if you could see it's actually following my instructions the AI agent so I'll load your voice skill first then write this email so boom like it's as if you're telling an an intern or someone on your team to get something done

[译文] [Rashid]: 这里应该发生的是，它应该通过邮件 Skill（我在这里打开了这个），如果你看这里的邮件 Skill，在 Skill 内部我有“读取 Rashid Voice Skill”，所以它路由到了这里。如果你能看到的话，AI 智能体实际上正在遵循我的指令：“我会先加载你的 Voice Skill，然后写这封邮件。”Boom，这就像你在告诉实习生或团队里的人去完成某件事。

[原文] [Rashid]: now what's even better is that this cloud code is connected to my notion database so I can ask it to do that to move it to my notion database... yes I want you to save it to notion email database and I want you to turn this into a uh quote post

[译文] [Rashid]: 更棒的是，这个 Claude Code 连接到了我的 Notion 数据库，所以我可以要求它把内容移到我的 Notion 数据库里……是的，我要你把它保存到 Notion 邮件数据库，并且我要你把它变成一个引用语帖子（Quote Post）。

[原文] [Rashid]: so I have the skill it went through the skill over here social media workflows create paragraph code so social media skill it went through that and now it's spawning you to save the email to notion okay so that agent's doing that so I'm going to talk to our main agent now i want paragraph looks good can you um create it

[译文] [Rashid]: 我有这个 Skill，它通过了这里的社媒工作流，创建段落代码，也就是社媒 Skill，它通过了那个。现在它正在提示保存邮件到 Notion。好的，那个智能体正在做这个。我现在要跟主智能体说话：“段落看起来不错，你能创建它吗？”

[原文] [Rashid]: so now it's using the notion API it's going to find the database and then it's going to create that email for us okay so it just added it to notion now it's going to create the uh paragraph code carousel for us

[译文] [Rashid]: 现在它正在使用 Notion API，它会找到数据库，然后为我们创建那封邮件。好的，它刚刚把它加到了 Notion 里。现在它要为我们创建段落代码轮播图（Carousel）。

[原文] [Rashid]: so now it's creating um this first slide it's using Nanab Banana Pro and then it's going to create the second slide doing this live is going to be interesting how this comes out so it took the prompt so it it even created the prompt to send it here so now it's creating the visual it's using the API that I gave it access to man this is this is crazy man like ju just me like thinking about this like it would take me like at least 30 minutes to get this done and this is getting done automatically

[译文] [Rashid]: 现在它正在创建第一张幻灯片，它用的是 Nanab Banana Pro（可能是指特定的绘图模型或工具），然后它要创建第二张幻灯片。现场做这个很有趣，看结果如何。它提取了提示词，甚至创建了发送到这里的提示词。现在它正在创建视觉图像，它正在使用我给它访问权限的 API。天啊，这太疯狂了，光是想想这个，以前我至少要花 30 分钟才能做完，而现在这是自动完成的。

以下为您整理的章节 7内容。这是全文的最后一个章节，重点讲述了 Skill 的复利效应与对未来工作的展望。

章节 7：复利效应与未来展望 (The Compounding Effect & Future of Work)

📝 本节摘要：

在最后的总结中，Rashid 提出了 Skill 的“剧本链”（Playbook Chain）概念，即通过链式反应将多个 Skill 串联（如：邮件生成 -> 社媒转换 -> 自动发布 -> 数据反馈），从而构建出能像员工一样工作的自主闭环系统。他算了一笔账：仅自动化一个耗时 30 分钟的任务，每年就能节省约 3 个工作周的时间。最后，他展望了未来（2026年），认为通过不断堆叠这些 Skill，企业主将能让 AI 智能体承担 80%-90% 的实际工作，实现真正的业务规模化。

[原文] [Rashid]: and the real powerful thing about skills is that it compounds so think of it like a playbook chain so if you have an agent and you have multiple skills that you want it to do it is able to do that within cloud code

[译文] [Rashid]: 而 Skills 真正强大的地方在于它会产生复利效应（Compounds），所以把它想象成一个“剧本链”（Playbook Chain）。如果你有一个智能体，并且你有多个希望它执行的 Skill，它能够在 Claude Code 中做到这一点。

[原文] [Rashid]: so let's say I have a skill A which is I want it to help me create an output let's say this is an email that's skill A skill B could be like transforming it i could turn this into a social post and then social C could be like publishing it like a skill on how to publish it and then skill D could be like going through your social media and then extracting the results on what happened to that spec specific post

[译文] [Rashid]: 比方说我有 Skill A，我希望它帮我创建一个输出，假设这是一封邮件，这是 Skill A；Skill B 可以是转化它，我可以把它变成一个社交媒体帖子；然后 Skill C（此处口误，应为 Skill C）可以是发布它，比如一个关于如何发布的 Skill；然后 Skill D 可以是浏览你的社交媒体，然后提取关于那篇特定帖子的数据结果。

[原文] [Rashid]: so then this can create a feedback loop where you can have agents that run with skills that know how to do things just like how employees would work in your business

[译文] [Rashid]: 这样这就创造了一个反馈循环，在这个循环里，你可以拥有运行着各种 Skill 的智能体，它们知道如何做事，就像员工在你的企业里工作一样。

[原文] [Rashid]: so this is super powerful because now you can have AI agents that are able to do actual work in your business but the only thing that they depend on is having clear skills that you know how to teach them how to use that

[译文] [Rashid]: 这超级强大，因为现在你可以拥有能够为你企业做实际工作的 AI 智能体，但它们唯一依赖的是拥有清晰的 Skill，也就是你知道如何教它们去使用的那些技能。

[原文] [Rashid]: so again here are some other examples so let's say you're a consultant you have an intake form you could turn that into proposal that could be turned on into onboarding emails and then you can have uh reporting on how the onboarding went

[译文] [Rashid]: 再举一些其他的例子。假设你是一名顾问，你有一份需求表单（Intake Form），你可以把它变成提案，提案可以变成入职邮件，然后你可以生成关于入职情况的报告。

[原文] [Rashid]: let's say you're an agency you get a brief from a client that creates a deliverable then it creates a review then an invoice like so on so these skills can connect like daisy chains you can have as many skills as you want you can have them route between other skills so this becomes very very very powerful

[译文] [Rashid]: 假设你是一家代理商，你从客户那里收到一份简报，它生成一个交付物，然后生成一份审查，然后是一张发票，以此类推。所以这些 Skill 可以像“菊花链”（Daisy Chains，指首尾相连）一样连接起来。你想要多少 Skill 都可以，你可以让它们在其他 Skill 之间路由，所以这变得非常非常非常强大。

[原文] [Rashid]: so if you just think about it if you are spending 30 minutes doing a specific task and you get a skill to do that within 3 minutes that's going to save you 2.5 hours a week and then that adds up to 10 hours a month and then if we extend that to a year that's 120 hours a year and that is equivalent to three work weeks

[译文] [Rashid]: 你试想一下，如果你花 30 分钟做一个特定的任务，而你用一个 Skill 在 3 分钟内完成它，那每周将为你节省 2.5 小时；累加起来就是每月 10 小时；如果我们把它延伸到一年，那就是每年 120 小时，这相当于三个工作周。

[原文] [Rashid]: so imagine just implementing one skill will save you three work weeks and that one skill can compound to more you can build more skills so 1 2 3 four five skills

[译文] [Rashid]: 想象一下，仅仅实施一个 Skill 就能为你节省三个工作周，而且这一个 Skill 还能复利增长成更多，你可以构建更多 Skill，比如 1 个、2 个、3 个、4 个、5 个 Skill。

[原文] [Rashid]: so now we're reaching a point where we're reaching a point in business where you can pretty much have AI agents do 80 to 90% of the work in your business

[译文] [Rashid]: 所以现在我们正达到一个临界点，在商业中我们正达到这样一个点：你可以让 AI 智能体完成你业务中 80% 到 90% 的工作。

[原文] [Rashid]: so if you actually want that subscribe to this channel and pay attention to my next videos because I'm going to be sharing more insights on how you could get AI agents and digital assets to do majority of the work for you in your business and scale in 2026 see you in the next video

[译文] [Rashid]: 所以如果你真的想要那样，请订阅这个频道并关注我接下来的视频，因为我将分享更多关于如何让 AI 智能体和数字资产为你完成业务中的大部分工作，并在 2026 年实现规模化的见解。下个视频见。

**

这份文档通过逻辑分层、摘要提炼与双语对照，完整复原了 Claude Skills 从概念原理、Web 端基础操作到 Code 端高级自动化的全貌。希望这份资料能为您构建自己的 AI 工作流提供有力支持！