Claude Skills for Financial Advisors: A Practical Guide to Building and Using Your First One

A Skill is a piece of capability you install once into Claude. After that, Claude knows how to do a specific thing for you the same way whenever you ask.

That's the whole concept.

To understand why that matters, think about how you've used AI up to now. Whenever you open Claude or ChatGPT, you start from scratch. You write a new prompt, paste in context, and explain what you want. The conversation ends and the next time you open the app, you do all of that again. Nothing compounds, and every Monday morning is Monday morning all over again.

A Skill ends that. The work you do once to set it up becomes the way your AI works permanently. You type a short command (something like "train me on my voice" or "scan my niche") and the Skill takes over. It walks you through what it needs, runs the task, and saves the output to a folder on your computer.

Skills live inside Cowork, the desktop application Anthropic ships with Claude. If you haven't set up Cowork yet, we cover that in the Claude AI Practical Guide. The short version is that Cowork is the workspace where Claude can access files on your computer, run multi-step tasks, and remember context across sessions. Skills are the building blocks you add to that workspace.

A few things Skills are not. They aren't autonomous agents that make decisions for you, they aren't connected to your CRM or your custodian or your client data unless you point them there (and you shouldn't), and they aren't a replacement for your voice or your judgment. They're a way to do one specific thing well, repeatedly, with the compliance language built in from the start.

That definition matters for any professional. For advisors specifically, three things make Skills more valuable than they are for almost any other field.

Voice preservation

AI without a voice guide produces generic content. You've seen it, and everyone has. The em dashes, the bullet points that don't sound like you, and the formal language that nobody actually writes in. It's serviceable, and it's also recognizable as AI to anyone reading it, especially your clients.

AI with a voice guide produces better content, but it requires you to attach the guide whenever you start a new conversation. Many advisors forget. Even when they don't forget, they end up with slightly different outputs across different chats because they didn't structure the guide the same way each time.

A Voice Skill carries your voice instructions with it whenever you call it, without you having to remember anything. The result is that your AI sounds like you in every output, not just the ones where you set up the prompt correctly. The compounding finally starts.

Compliance consistency

Every prompt an advisor uses should have compliance guardrails. Frame as education and not advice, use no client names, write no specific recommendations as directives, and offer no guarantees or forward-looking promises.

In practice, advisors forget. Or they remember sometimes and not others. A Skill carries those guardrails permanently. Once you build a Skill that requires "frame as education not advice, no client identifying information, all output as my own reflection rather than directive," every output it produces respects those rules. You don't have to remember, the Skill remembers for you.

This is the part that separates a Skill from a clever prompt you saved in a Google Doc. A prompt requires you to copy and paste it, with all of its guardrails, whenever you use it. A Skill does it for you automatically.

Repeatability without resetting

The most expensive part of using AI as an advisor is starting from scratch every Monday. New conversation, new context, and new explanation of what you want. The cost shows up as friction, and friction is what stops people from using AI consistently.

A Skill ends the reset. You set it up once, and it works the same way for you every week, every month, and every year, until you decide to update it. The work compounds because the system is finally stable.

Skills aren't a productivity hack. They're the way an advisor's AI work finally compounds instead of resetting every week.

Let me show you what this looks like with a real example.

A few months ago I asked Claude to write a client email scheduling an annual review. It's a standard task, and every advisor sends this email.

I gave Claude a simple prompt.

"Write a client email to a couple letting them know we're scheduling their annual review next month and asking them to think about any life changes since we last met."

That's the whole prompt. No voice instructions, no context, and no examples, just the task.

Here's what Claude wrote.

Dear [First Name] and [First Name],

I hope this message finds you well. I'm writing to let you know that it's time for your annual financial review, and I'd like to get something scheduled for next month.

These reviews are most productive when we come prepared, so I'd encourage you both to take some time beforehand to reflect on anything that has changed since we last met. This might include:

• Changes in employment, income, or compensation structure

• Major financial decisions or purchases you're anticipating

• Updates to your family situation or estate planning needs

• Evolving thoughts on your retirement timeline or goals

• Any financial concerns or questions that have come up

There's no need to have everything sorted out ahead of time - part of what we'll do together is work through what any changes mean for your plan.

Please use the link below to select a time that works for your schedules, or feel free to reply directly and I'll coordinate from there. [Scheduling Link]

I look forward to connecting with you soon.

Warm regards,

[Your Name]

Read that and think about whether it sounds like you. For me it didn't. The "Dear" salutation, the bullet points (which I rarely use in client emails), the em dashes, and the formal "I hope this message finds you well" opener. It's a perfectly fine email, and it's also clearly written by an AI that doesn't know who I am.

Then I attached my voice guide and gave Claude the same exact prompt. Two sentences of instruction and one file. Here's what it produced.

Hi [First Name] and [First Name],

It's that time of year again, and I'm already looking forward to sitting down with you both.

We're scheduling annual reviews for next month, and I'd love to get something on the calendar that works for you. A phone call or video meeting works fine, whatever's easier.

Before we meet, I'd encourage you to think through the past year a little. Not in a stressful way, just a quiet inventory. Any big changes since we last talked? A job shift, a new goal, a family update, something that felt small at the time but might matter now? Those are exactly the things worth bringing.

As we discussed when we first started working together, the plan is really just a reflection of your life. When life changes, we want to make sure the plan keeps up.

Reply whenever you're ready and we'll get something on the books. And as always, don't hesitate to reach out if anything comes up before then.

Talk soon,

[Your Name]

That one sounds like me. The "Hi" instead of "Dear," the casual tone, the "whatever's easier," the "as we discussed when we first started working together" line that pulls the client back into a shared history, and the "don't hesitate to reach out" close. Even the rhythm of the questions about life changes feels like how I actually talk.

Same prompt, same task. Two sentences of instructions and a voice template made the difference between an email that sounds like every other advisor and an email that sounds like me.

That's what a Skill does, except you don't have to remember to attach the voice template. The Voice Training Skill builds the template once, saves it to your computer, and every future Skill that uses voice as an input pulls from it automatically. You write the email, the Skill makes sure it sounds like you, and you never have to think about it again.

The Voice Training Skill itself walks you through four questions, one at a time. Your audience, your areas of expertise, samples of writing you've actually done, and the things you never want AI to write. After fifteen minutes of conversation, you have a complete voice template saved as a file on your computer. From that point forward, you're never writing a client email from scratch in a generic AI voice again.

One compliance note that matters here. The Voice Skill does not store anything outside your own Claude-Work folder on your computer. It works on your writing samples, full stop. No client data ever touches it, and no identifying information ever leaves your machine. This is the bright line that makes Skills usable for advisors at all, and it's the line every Amplify Skill is designed around.

I teach financial advisors how to use AI for content, communication, and client attraction. New frameworks and prompts every Tuesday and Friday. Subscribe free or get full access for $20/month at amplifyforadvisors.substack.com.

The setup is smaller than many advisors expect.

• A Claude Pro subscription, $20 per month from Anthropic at claude.com

• The Claude Desktop app, free, from claude.com/download (Mac and Windows both work)

• A folder on your computer called Claude-Work (full setup details are in the Practical Guide)

• Cowork enabled inside the desktop app

• A handful of writing samples you wrote yourself (LinkedIn posts, client emails you've sent, blog drafts) for any voice-based Skill

• About fifteen minutes

That's everything. If you already have Claude Pro and the desktop app installed, you're a few clicks away from being ready to install your first Skill.

The hardest part isn't the install, it's deciding which Skill to build first. Here's how to think about that.

Rule one: compliance guardrails are non-negotiable

Every advisor Skill must include explicit compliance language. The Skill should refuse to write client recommendations as directives, refuse to use client names if they're pasted in, and frame all output as education rather than advice.

If a Skill you're considering doesn't have these guardrails built in, build them in before you start using it. Don't trust generic prompts written for general business audiences. The compliance environment for advisors is different from the compliance environment for marketers, and your Skills should know that from the first line.

Rule two: the Skill works on your inputs, not your clients' data

This is the bright line that makes Skills usable in our profession.

Every Amplify Skill is designed around one principle. You feed the Skill your writing samples, your niche, your topics, your authority content, your voice. You never feed it client account data, financial statements, planning documents, or personally identifiable information about anyone you work with.

The Skill stays on your side of the compliance line by design. If you ever feel uncertain about whether something belongs in a Skill, the default answer is to leave it out. The Skill is about making your work sound more like you, not about processing client data.

Rule three: voice comes before output

Build the Voice Training Skill first. Use the voice template it produces as the input to every other Skill you build.

This is the rule that separates advisors whose AI sounds like them from advisors whose AI sounds like every other advisor on LinkedIn. The Voice Skill is the foundation, and everything else is the building on top of it. If you try to build a Content Generator or a Newsletter Drafter or anything else before you have a voice template, the output will be generic. The Skill is only as good as the voice it's working from.

These three rules are the difference between a Skill that helps you build your practice and a Skill that creates a compliance problem you didn't see coming.

The Skills I build for Amplify follow a layered architecture. Each one is useful on its own, and they get more useful when you stack them.

Layer one: Voice Training. Builds your voice template. The foundation for everything else.

Layer two: Niche Authority Scanner. Reads your voice template and your niche, then surfaces the content topics your ideal clients are actually searching for. Tells you what to write about.

Layer three: Content Generator. Multi-platform content creation that pulls from your voice template and the topics the Scanner found. Writes drafts in your voice for LinkedIn, your newsletter, and your blog.

Layer four: Content Recycler. Reads your existing content library and finds the posts worth bringing back, rewritten in your current voice for your current audience.

Each Skill feeds the next. Voice template feeds the Scanner, the Scanner's output feeds the Generator, and the Generator's output feeds the Recycler. The system gets stronger as you use it.

All of this lives inside Claude but is accessible through the Amplify for Advisors subscription at twenty dollars a month, with new Skills shipping roughly monthly. Subscribers get every Skill, every prompt, and every framework as they're built.

This article tells you what Skills are and how they work in principle. The actual Skills, the prompts, and the templates live inside the subscription because they are the product. If you want the system ready-built, that's the path, and if you want to build something yourself, the next section gives you the starting point.

Two paths from here, depending on how you want to work.

Path A: Build your first Skill yourself

1. Sign up for Claude Pro at claude.com if you don't have it already.

2. Download Claude Desktop at claude.com/download and install it.

3. Set up your Claude-Work folder using the steps in the Practical Guide.

4. Use the voice guide prompt from Issue 2 of Amplify to build a voice template manually.

5. Save the voice template as a markdown file inside your Claude-Work folder. That's your first Skill input.

This path takes a few hours of focused work. It's real, and it works. Some advisors prefer to build their own systems from scratch, and if that's you, this is the road.

Follow these steps to build your own skill:

1. Enable the Skill Creator

• Open Claude (on the Web or the Desktop App).

• Go to Settings > Capabilities and make sure Code Execution and File Creation is turned on.

• Navigate to Customize > Skills.

• Click the "+" button and select "Create skill".

2. Have Claude Build It

• Choose the "Create with Claude" option.

• Tell Claude exactly what you want to automate. For example: "I want a skill that summarizes my weekly meeting transcripts, extracts action items, and formats them as a bulleted email draft."

• Claude will ask a few clarifying questions about your preferences and then draft the skill (the .md file) for you.

3. Save and Test Your Skill

• Once Claude finishes, save or export the package as instructed.

• Upload the ZIP or skill folder via the Customize > Skills menu.

• Test it out in a new chat by simply saying: "Use my [Skill Name] skill to..."

Path B: Subscribe and get the Skills ready-built

Subscribe to Amplify for Advisors at $20 per month. Get the Voice Training Skill, the Content Recycler, and every future Skill as they ship. Included in the subscription are the prompts, the frameworks, and the twice-weekly newsletter that teaches the thinking behind each Skill.

Many advisors who try Path A end up moving to Path B once they realize how much trial and error went into the Skills that are already built. Months of testing compressed into something you install in five minutes. That's the trade.

Either path is honest work. Pick the one that matches how you actually want to spend your time.

Claude shipped Skills as a feature in early 2026, KPMG made the partnership official in May, and Anthropic launched wealth management plugins for Cowork shortly after. The pace is accelerating, not slowing down, and the advisors who build the habit of using Skills now are going to compound on that habit for the next several years.

You don't need to be early to everything. You do need to be early to the things that change how you actually work. Skills are one of those things.

Start with voice. Build the template, save it to a folder on your computer, and use it as the input to everything else you build from here.

The work compounds from there.