Your First Claude Skill: Email Triage in One Prompt

In a previous video, Claude got access to your inbox, but every morning you're having to type the same prompt and paragraph of instructions. This one's pretty short, but over time it's probably going to become more and more robust as it goes to look at your email and do all the triage you want it to do. That was a simplified example. Today we're going to fix that permanently.You're going to build your first skill, and from now on the whole routine will be one prompt.So what is a skill, actually? Let's go under Customize and we'll see Skills. I can create skills through the skill creator, but let's talk about this first. A skill is instruction files for Claude to complete a specific task in a repeatable way. We use them to encode our personal best practices into a repeatable set of instructions.These aren't crazy programming files. They're really just text files telling Claude what to do. Some of them have a few other things in them, but these are written in plain English, and the skill is written out just like any large prompt would be.We're going to use the skill creator to create other skills. It's basically a checklist in plain English: a name, a description, when to use it, and the steps. Think of it like a standard operating procedure, or onboarding an assistant.A prompt is a request. A skill is a procedure. Steps you repeat, procedures you write once. Most people stay stuck at requests forever. What we're doing today bridges that gap.For your first skill, pick something you've already asked Claude to do three or more times. Think email triage, morning meeting prep, or weekly reporting. We're doing email triage because it stacks on top of the connector from the last video. Build on what's already wired. Each layer should use the last one.So let's create a skill, and we're going to create it with Claude. I'll say: "Let's create a skill together using your skill creator skill. First, ask me what the skill should do."I'm going to be using Sonnet here. We could upgrade to Opus, but not on my paid plan account. If I were on a higher tier, I would. I think Sonnet is fine for what we're doing here.It asks: "What would you like the skill to do?" I'll say: summarize my unread email from the last 24 hours and put it into three buckets. Then I'll add to it: if something is high priority, go ahead and write a draft response.It says, "Great concept." Then a browser popup asks if I want to be notified. Sure, that's fine, that's a different problem.So it's going to go create this workflow using the built-in skill creator. It creates the skill file, the name, the description, the trigger phrases, and everything necessary to make the skill run without us having to do anything besides have a conversation with this interface. It's the same way we'd do this with a remote or virtual assistant who was skilled in these tasks.So now we've got this email triage skill. It can triage into three buckets, present a summary, draft replies, and give a final summary. We could edit it and keep going back and forth with Claude to get the skill exactly how we want.Here's what it does, step by step: it fetches your unread email from the last 24 hours, triages it into needs reply, can wait, or ignore, summarizes it, auto-drafts replies in a professional tone saved directly to your Gmail drafts, and reports on every draft it creates.It asks if I want to test it now. I'm going to go ahead and save the skill instead. And it's saved.Here are a couple of interesting things. I can trigger this by saying "go summarize my inbox." The triggers include "what emails do I have," "summarize my inbox," and "catch me up on email." But I can also just type a forward slash and the skill name. I'll type forward slash, and there's email triage. When I hit enter, it runs that skill.So whether I use the slash command, or type "what emails do I have," or any of the other triggers like "triage my email," it runs. And it's done it again: your inbox is essentially clear. Now I can run this skill any time I want.If I go to Customize and then Skills, I'll see the email triage skill has been saved. This is also where I can edit it. If I want to add more to it later, I can.So we can iterate on this. For example, I could have it handle newsletters differently. I'll say: update the email triage skill so newsletters get one summary line instead of being listed individually. We didn't have any newsletters here, but I know I'll have them in the future.Now it goes in, reads the skill, copies it, and makes the updates it needs to. This is how we iterate. Just like with a remote assistant, we'd email or text them and tell them what to do differently the next day.It's like a new hire's first week. We don't fire them for formatting something differently or putting newsletters in the wrong place. You give feedback until it sticks. So here I'm giving update feedback: in the future, whenever we get newsletters, put them all on one line. I don't want to know about them individually.This one skill is really a convenience, but a library of these becomes a system. You get triage, then meeting prep, then proposal formatting, then project research. Every procedure you capture into a skill is something you can leverage and compound.So it's updated the skill. It came in and said, okay, whenever there are newsletters, we'll handle them differently and present the summary right here. We can see where it updated the newsletter handling. It made the changes we asked for, and now we're going to save the skill. It asks if we want to replace it. Yes, update and replace.It was that simple. No code, no changes. And the kicker is something we'll uncover soon: a skill can run on a schedule. This triage we just built can happen at 6 a.m. every day without us. That's where we're headed.So, a couple of assignments. Build the triage skill today using the package, exactly as we showed it. Then by the end of the week, capture one more workflow you've repeated at least three times that should be done by Claude and not by you. Don't overthink which one. The habit of capturing matters more than the one you pick.In the next video, we're going to talk about securing Claude so we're not sharing this information with Anthropic.