Claude Skills for Financial Advisors: 8 Compliance-Aware Workflows (2026)

Claude Skills for financial advisors are reusable capability packages that let an RIA's Claude workspace produce the same client-ready output every time — meeting recaps, Investment Policy Statements, quarterly letters — without re-prompting. Anthropic released Skills in October 2025 as folders containing a SKILL.md file plus optional scripts and reference documents. Claude only loads a Skill when the task matches its description, so a firm can keep dozens of them installed without slowing anything down.

For a wealth-management firm, Skills are the fastest way to encode the CCO's compliance guardrails and the lead advisor's writing voice into every output. They work across Claude.ai, Claude Code, Claude for Small Business, and the Anthropic API — so the same recipe your associate uses in the browser is the one your back office uses through the API.

This guide gives you 8 Skill recipes tuned to the SEC Marketing Rule (Rule 206(4)-1) and the FINRA Regulatory Notice on generative AI. Each one names the trigger, the guardrails, and the reference files a compliant Skill needs — so your first Skill is one your CCO will actually approve.

What Claude Skills Are (And Why RIAs Should Care)

A Claude Skill is a folder containing a SKILL.md file with YAML frontmatter, plus optional .py scripts and reference documents that Claude loads on demand. The frontmatter tells Claude when to activate the Skill — for example, "use when drafting a quarterly client letter." Claude only reads the full Skill contents once a task matches, which keeps token cost low even with a large firm-wide library.

For an RIA, this matters because compliance is repetitive. Every quarterly letter needs the same disclosures. Every prospect research brief needs the same conflict-check step. Every meeting recap needs the same "not personalized advice" line. A Skill encodes those rules once, then applies them to every future run.

Skills are portable across Claude.ai, Claude Code, Claude for Small Business, the Anthropic API, and Enterprise. That means the recipe your associate advisor uses in the browser is the same one your ops team runs through the API for batch quarterly letters — no divergent prompts, no drift between "the compliant version" and "the fast version."

• A Skill = SKILL.md (with YAML frontmatter) + optional scripts + optional reference docs (disclosures, letter templates, IPS shells)
• Claude only loads the full Skill when a task matches — dozens of Skills add ~100 tokens of overhead each
• Skills run across Claude.ai, Claude Code, Claude for Small Business, and the Anthropic API — one recipe, every surface
• Different from a Custom GPT (competitor runtime) and different from a Claude Project (static context, not reusable)
• Lighter than an MCP server — no infra to run, no live data connections; Skills are packaged how-to's, not live tools

The Compliance Guardrails Every Advisor Skill Needs

Every Claude Skill an RIA deploys should be built against the SEC Marketing Rule and FINRA's generative-AI guidance before it ever touches a client-facing document. The SEC's 2024 and 2025 risk alerts on advisor use of AI have repeatedly flagged unsubstantiated performance claims, undisclosed hypotheticals, and "AI washing" language in marketing. FINRA Regulatory Notice 24-09 reminded firms that existing rules — supervision, recordkeeping, communications with the public — apply to GenAI output the same way they apply to a human-drafted email.

That means a compliant advisor Skill has three non-negotiable parts. First, the SKILL.md must name the review step: "output must be reviewed by a registered supervisor before delivery." Second, the reference files must include your firm's current disclosure library — not sample text pulled from the web. Third, every draft the Skill produces must be preserved under Rule 17a-4 recordkeeping, which means the output goes into your compliance archive, not just an inbox.

A useful shortcut: treat every Skill output as a Rule 206(4)-1 "advertisement" until proven otherwise. If the draft references performance, references a client testimonial, or could be sent to more than one prospect, it is an advertisement and needs the fair-and-balanced treatment the rule requires.

• Bake the current firm ADV Part 2A disclosure into a disclosures.md reference file the Skill loads on every run
• Add a "supervisor review required" line in the Skill's YAML description so Claude always surfaces it
• Never let a Skill generate hypothetical or projected performance numbers — hard-code a refusal in the SKILL.md
• Route every output to your recordkeeping archive (SMARSH, Global Relay, or your CRM) — not just email
• Version-control your Skills folder in Git — the CCO needs to answer "what did the AI say on March 14?" a year later

Skill 1: Compliant Client Meeting Recap

The Meeting Recap Skill turns a transcript or advisor notes into a client-ready follow-up email that never crosses into personalized advice unless the underlying meeting authorized it. The Skill's frontmatter triggers on phrases like "recap the client meeting" or "draft a follow-up." Its reference files include your firm's recap template, the "no new advice" disclaimer, and a list of action items the recap should always confirm (next meeting, documents needed, rebalance timing).

Behavior the Skill enforces: it never introduces investment recommendations that were not discussed in the source notes. It never quantifies expected returns. It ends with the firm's standard "please review and reply with corrections" line, which creates a client-confirmed record for compliance.

A useful failure mode to guard against: without a Skill, Claude will happily fabricate a "we agreed to rebalance into a 60/40" line if the notes are ambiguous. The Skill's system prompt should explicitly instruct: "if a decision is not explicitly recorded, list it as an open question, not a decision."

• Trigger phrase: "draft a meeting recap" or "client follow-up from today's call"
• Reference files: recap template, firm disclosure block, "open questions vs. decisions" rule
• Output section: Attendees / Discussed / Decisions / Open Questions / Next Steps — with dates
• Hard rule in SKILL.md: no new advice, no performance projections, no quantified outcomes
• Route the draft to the advisor's inbox with a "supervisor review required before send" flag

Skills 2 & 3: IPS Drafts and Quarterly Performance Letters

The Investment Policy Statement (IPS) Skill and the Quarterly Letter Skill are the two highest-leverage recipes for an RIA, because both are repetitive, template-driven, and expensive when a paraplanner writes them from scratch. The IPS Skill takes the intake questionnaire (risk tolerance, time horizon, liquidity needs, constraints) and produces a draft IPS in your firm's exact section order. The Quarterly Letter Skill takes the client's account performance data and produces a compliant letter with the required benchmarks, disclosures, and net-of-fee returns.

For the Quarterly Letter Skill specifically, the SEC Marketing Rule's performance requirements are the guardrail. Gross returns must always be shown alongside net-of-fee returns. Time-weighted returns need clear labeling. Any benchmark must match the strategy — a diversified balanced portfolio does not get compared to the S&P 500 alone. Your Skill's performance-rules.md reference file should encode all of this.

For the IPS Skill, the guardrail is the "no advice without a signed agreement" line. The Skill produces a draft — not a final IPS — and the advisor signs off before it goes to the client. Layer3 Labs typically builds these with a draft- filename prefix that makes it visually obvious the document has not been reviewed yet.

• IPS Skill: sections in firm order — Objectives / Constraints / Asset Allocation / Rebalancing / Review Schedule
• Quarterly Letter Skill: always show gross AND net-of-fee, match benchmarks to strategy, label time-weighted returns
• Both Skills load your firm's current disclosure block from disclosures-2026-q3.md — update quarterly
• Both Skills produce drafts with a draft- filename prefix — the human review step is unmissable
• Batch mode: the Quarterly Letter Skill can run against 200 accounts through the API in one job

Skills 4 & 5: Tax-Loss Harvesting Narratives and KYC/CIP Summaries

The Tax-Loss Harvesting (TLH) Narrative Skill and the KYC/CIP Summary Skill both take structured data and produce readable, review-ready prose. The TLH Skill takes a trade blotter — lots sold, replacement securities, realized losses, wash-sale checks — and produces the client-facing explanation of what happened and why. The KYC Skill takes the client's onboarding intake (identification documents, source-of-funds notes, PEP screen, sanctions check) and produces a summary the CCO can review before account opening.

The TLH Skill should never claim tax savings without the disclosure that individual tax outcomes depend on the client's full return — a line the IRS and FINRA both watch for. Your Skill's reference file should list the specific phrasings your CCO has pre-cleared. The Skill picks from the pre-cleared list rather than generating fresh disclosure language.

The KYC Skill is more of a summarizer than a writer. Its job is to say "here is what we have, here is what is missing." Give it a checklist reference file — the exact CIP items your firm requires — and instruct it to flag gaps rather than fill them in. A Skill that invents a source-of-funds explanation is a supervisory failure waiting to happen.

• TLH Skill: input is the trade blotter, output is the client-facing explanation + wash-sale confirmation
• TLH Skill hard rule: pick disclosure language from the pre-cleared list only, never generate new phrasing
• KYC Skill: summarizes what you HAVE — flags what is MISSING — never fills in unverified fields
• Both Skills preserve full source data alongside the draft for audit trail (Rule 17a-4)
• Route KYC Skill output to the CCO's review queue before account opening, not to the client

Skills 6, 7 & 8: Prospect Research, Fee Explainers, and Portfolio Review Prep

The final three Skills round out the advisor's weekly rhythm: prospect research briefs, fee-disclosure explainers, and portfolio review prep. The Prospect Research Skill takes a name, firm, and public role and produces a briefing note before a discovery call — public bio, likely financial priorities based on career stage, and topics to avoid. Because this touches on non-public inference, the Skill's guardrail is "only cite publicly available sources with URLs."

The Fee Explainer Skill takes the firm's fee schedule and a specific client's AUM tier and produces the plain-English explanation the client will read in the ADV Part 2A follow-up. The regulatory guardrail is the "all-in" cost obligation — the Skill must include underlying fund expense ratios, custodian fees, and any performance fees, not just the advisory fee. The 2026 SEC exam priorities named fee-clarity as a continuing focus area.

The Portfolio Review Prep Skill takes the client's current holdings and produces the advisor's prep document for the annual review — drift from target allocation, rebalancing candidates, tax-loss opportunities, RMD status. This one stays internal. The Skill's hard rule is "advisor prep only — never send to client." A for-internal-use-only watermark in the output template makes that unmissable.

• Prospect Research Skill: public sources only, cite URLs, flag any inferred (non-cited) claim as "to verify"
• Fee Explainer Skill: always show all-in cost — advisory fee + fund expenses + custodian fee + any performance fee
• Portfolio Review Prep Skill: internal-only, watermark, never routed to client email
• All three read the same firm-wide disclosures.md reference file — one source of truth, updated centrally
• Portfolio Review Skill pairs well with a Claude Code MCP connection to your portfolio management system

How to Deploy These Skills Across Your Firm

Deploy Claude Skills at the firm level by storing them in a shared Git repository and installing them into every advisor's Claude workspace at once. The Skills folder pattern Anthropic published — each Skill is a directory with SKILL.md at the root — means version control and pull-request review work naturally. Your CCO reviews Skill changes the same way a developer reviews code changes: through a diff.

For a small RIA, the practical rollout is three weeks. Week 1: install Claude for Small Business or Enterprise, set up the shared Skills repo, and load the Meeting Recap and Quarterly Letter Skills. Week 2: add IPS, TLH, and Fee Explainer Skills. Week 3: add KYC, Prospect Research, and Portfolio Review Prep. Every Skill goes through CCO review before it goes live.

The failure mode to avoid: individual advisors building "shadow Skills" in their personal Claude accounts. If it is not in the firm's repository, it is not supervised, and its output is not preserved under Rule 17a-4. Layer3 Labs builds a "no shadow Skills" policy into the CCO's written supervisory procedures during onboarding.

• Store the Skills folder in a private Git repo — the CCO gets pull-request review rights
• Version every disclosure file by quarter (disclosures-2026-q3.md) — never edit in place
• Enable Claude for Small Business or Enterprise so Skills apply firm-wide, not per-user
• Write "no personal Skills" into the firm's written supervisory procedures
• Log every Skill output to your recordkeeping archive (SMARSH, Global Relay, MyRepChat, etc.)

What you need to run Claude Skills for financial advisors

The first question most financial advisors teams ask is whether their current setup can handle Claude Skills. For the standard cloud version, the answer is usually yes: Claude Skills runs on the provider's servers, so the computers and internet connection you already have are enough to start — there is no server to buy and nothing to install across the firm.

What you do need is two things: access (a business plan or the API) and a tool to work in. Whoever wires Claude Skills into your workflows will move fastest inside an AI IDE — Cursor is the most popular and connects to Claude Skills directly — while the rest of the team uses Claude Skills's own apps day to day.

The exception is compliance. If client financial data and regulatory recordkeeping mean client data cannot leave your systems, the cloud version is off the table and you move to a private, on-prem setup: self-hosting an open-weights model on hardware you control. In practice that is a workstation with a strong GPU (an NVIDIA RTX 4090 build) or a large-memory Mac Studio for mid-size models, or RunPod to rent the same power by the hour. Our open-weights models for business guide walks through the full build.