Claude Skills for Private Equity: What Anthropic's PE Plugin Does and How to Make It Yours

1. The Short Answer: Start With Anthropic's Plugin, Then Make It Yours

Claude has private equity skills, and the place to start is Anthropic's own private equity plugin. It adds ten commands, each backed by a skill, for sourcing, screening, diligence, IC memos, returns, unit economics, portfolio reviews, value creation plans and an AI readiness scan.

The plugin knows how a generic private equity firm works. It has never seen your screening criteria, IC memo template or return hurdles, which are what make a first draft usable at your firm.

So install it to learn what each command does. Then fork it into a private repository, rewrite the skills around your templates, and push that copy to everyone as Required.

If you came looking for prompts, the commands are the better version: a prompt written down once, named and saved where the whole firm can run it.

The plugin gives you the outline of a deal process. The judgment that fills it in has to come from your firm.

2. Skills, Plugins, Projects and Prompts

Four words get used loosely. Here they are, from smallest to largest.

• A prompt is what someone types. It lives in one chat, and the good ones tend to stay in one person's notes.
• A project is a workspace for one piece of work, such as one deal or one fund, with its files, instructions and chats in one place.
• A skill is a method: a folder of instructions, scripts and resources that Claude loads when a task calls for it. Anthropic's developer docs describe each one as a directory with a SKILL.md file.
• A plugin is a bundle that installs as one unit: commands, the skills behind them and connector settings. Anthropic's private equity plugin is one.

Skills are the piece that scales. "Organization-wide skill management is available to Team and Enterprise plans," according to Anthropic's Help Center. An owner uploads a .zip containing SKILL.md in Organization settings, Plugins and skills, and "The skill is immediately provisioned to all users in your organization." Giving a skill only to specific groups, bundled in a plugin, is Enterprise only.

So use a project for anything tied to one deal, and a skill for anything your firm does the same way on every deal, such as a screening note or an IC memo. A prompt helps one person once. A skill helps everyone, on every deal after.

3. What Anthropic's Private Equity Plugin Includes

The plugin lives in Anthropic's financial-services repository on GitHub, in the folder plugins/vertical-plugins/private-equity. It is version 0.1.2, under the Apache-2.0 license, and its manifest describes it as "Private equity deal sourcing and workflow tools: company discovery, CRM integration, and founder outreach".

The private equity folder last changed on May 21, 2026, and the repository on September 21, 2026. A version below 1.0 and a folder untouched since May point the same way: treat it as a starting template your firm will own.

The plugin brings commands but no data. Its own connector file is empty, and all twelve data connectors sit in the companion financial-analysis plugin: Daloopa, Morningstar, S&P Global, FactSet, Moody's, MT Newswires, Aiera, LSEG, PitchBook, Chronograph, Egnyte and Box. In the README's words, "All connectors are centralized in the financial-analysis core plugin and shared across the rest," and "MCP access may require a subscription or API key from the provider."

So install financial-analysis first, then the private equity plugin. In Claude, open Customize, then Plugins, then Discover, then Add; in Cowork, open the Cowork tab first (Anthropic's plugin guide). For the whole firm, an organization owner adds Anthropic's marketplace in Organization settings, Plugins and skills, Marketplaces.

Two gaps are worth knowing. Anthropic's financial-services repository has no private credit or lending plugin as of October 23, 2026, so credit funds write their own skills. And none of the ten commands is a compliance check, so your data rules and review steps go into your own copy (section 9).

Install it for a test group and run every command once on a redacted past deal. An hour of that teaches more than any description, this one included.

4. The Ten Commands, One by One

Each command runs a skill. The middle column is Anthropic's description; the last is the part only your firm can write.

Read the last column as a to-do list: documents your firm already has and calls partners already make, which nobody has written down for Claude yet.

5. What the Plugin Cannot Know About Your Firm

A generic first draft is an expensive gift. Someone senior still rewrites it into house format, and that hour is the one the plugin was meant to save. The plugin cannot know these until you write them in:

• Your IC memo template. Sections, order, length, and whether the recommendation comes first or last.
• Your screening criteria. Size, sector, geography, margin floors and the deal breakers that end a conversation on page one.
• Your return hurdles. Gross and net targets, the holding period you underwrite, and the sensitivities your IC asks for.
• Your sector theses. Where you want to own businesses and why, which turns a sourcing list into your sourcing list.
• Your house style. How you write numbers, risks and mitigants, and the phrases partners strike from every draft.
• Your data rules. Which documents may go into Claude, what counts as material non-public information, and what never leaves the data room.

Most of this lives in a partner's head, an old Word template and the comments on last year's memos. Writing it down pays off before Claude reads a word, because it may be the first time anyone has had to agree on it.

Every firm that installs the plugin gets the same ten commands. These six items make yours different.

6. How to Make It Yours: Fork, Edit, Distribute

The plugin is open source, so you can copy it, change it and run your copy instead:

1. Fork it into a private repository. It will hold your criteria and templates, and an organization marketplace requires it: "public repos aren't allowed for organization marketplaces," per Anthropic's plugin admin page.
2. Edit the skills you will use. Rewrite each SKILL.md around the table's last column, add your templates to the skill's folder, and remove commands nobody will run.
3. Sync it as the organization marketplace, in Organization settings, Plugins and skills.
4. Set the install preference to Required. The choices are Installed by default, Available to install, Required and Not available. Set Anthropic's original to Not available, so nobody runs both.
5. Turn off User-created skills, so the firm's copy stays the one approved source.
6. Version every change. Record that you forked from 0.1.2, number each release, note who approved it, and bring in Anthropic's later updates on purpose.

After the fork, a change to how your firm writes a memo becomes a dated edit that someone approved. Few Word templates can say the same.

7. A Starter Skill: Your IC Memo in Your Template

Start with the IC memo. Every deal that reaches committee needs one, your template exists, and years of past memos make testing easy.

The skill is a folder holding SKILL.md, a plain-text copy of your template and one redacted past memo as an example. Here is an illustration, simplified: a real SKILL.md wraps the name and description in front matter, in the format Anthropic's developer docs show.

name: ic-memo
description: Drafts an investment committee memo in our house
  template. Use when someone asks for an IC memo or an IC draft.

Instructions
1. Follow the sections and their order in template.md exactly.
2. Use only the files provided. Cite the file and the page or tab
   for every figure.
3. Mark any figure you cannot trace with [CHECK].
4. Put the thesis in three sentences at the top.
5. Give each key risk a mitigant, and always show the downside case.
6. Leave the recommendation blank for the deal team to write.

Test it on two past deals whose final memos you still have. Give Claude the files the deal team had, compare the draft with what went to committee, and tighten the instructions until the format stops changing.

To hand it to the firm, an owner can zip the folder and upload it in Organization settings, Plugins and skills, or you can add it to your fork so it goes through the same review as the other commands.

The test that matters: show the IC chair two memos and ask which one started in Claude. Every figure in both still gets checked against the model and the data room.

8. Connectors: Where the Data Comes From

A skill is only as good as what it reads. Data arrives three ways:

• Licensed data providers. FactSet, PitchBook, S&P Global and the rest of the financial-analysis list connect through MCP servers, where your subscriptions allow.
• Your data room and document stores. Box and Egnyte are on the list. For other systems a connector can be built on the same protocol, as the MCP guide explains.
• Files people upload, kept in one project per deal.

Give each connector the least access that works. An owner enables a connector for the organization, and each person still authenticates individually. Owners set each connector's read-only and write/delete tools to Always allow, Needs approval or Blocked, organization-wide (Anthropic's connector article). Start read-only, because a screening skill only needs to read.

In Cowork, "Automatically approve" mode is on by default and "Always allow" for connector tools is off by default. For deal teams, switch the first off and leave the second off. Anthropic's Cowork safety page adds: "Avoid granting access to local files with sensitive information, like financial documents."

Connect the sources your analysts already trust, and nothing else. Every connector is one more vendor in your diligence file.

9. Governance for Skills

Once a skill drafts IC memos for the whole firm, give it the care you give the LBO model template.

• Who can publish? Owners provision organization-wide skills. Turn off User-created skills, give two or three people write access to the private repository, and route every change through a reviewed pull request.
• Who reviews? The partner who owns the template checks the content. The CCO checks any skill that touches material non-public information, LP data or personal data. Every release runs on a redacted past deal first.
• What backs the rules up? A rule in a skill is an instruction Claude follows. The hard limits are settings: the plan, connector permissions and retention.

The plan matters most. Anthropic's Privacy Center says: "By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models." The exception is thumbs up or down feedback, which stores the whole conversation for up to 5 years, so Team and Enterprise owners should switch off chat ratings in Organization settings, Data and Privacy.

Pro and Max run on consumer terms, with no organization admin console, no audit logs and no Compliance API, so skills and deal files belong only in the firm's Team or Enterprise organization. Git history records who changed each skill. Enterprise audit logs add 180 days of events, without chat content, and Team and Enterprise usage analytics show each skill's uses and cost per use.

Your Excel models already live by a simple standard: one owner, one current version and a record of every change. Hold your skills to it too.

10. Build It Yourself or Get Help

Many firms can do this alone. The editing is quick. Getting partners to agree on criteria and templates is slow, and no outsider can do it for you. You need four kinds of skill:

• Someone who knows the work, usually a VP or principal who has written the memos.
• Someone who writes clear instructions, since a SKILL.md is a precise brief for a capable new hire.
• Someone comfortable in GitHub, for the fork, the pull requests and the versions.
• An owner of the Claude organization, for marketplaces, install preferences and connector permissions.

At a small firm one person can cover three of those, if you protect a few of their hours each week. Skills written between deals compete with the deals.

Outside help earns its fee when nobody can own the work, when several strategies use different templates, when data room connectors need careful permissions, or when you also need the plan, settings and governance documents an LP DDQ asks for. The guide to choosing a Claude implementation partner covers how to pick one, and the 30-day adoption fix covers firms whose seats sit idle.

Either way, the partner who owns the IC template has to take part. A skill can only carry judgment your firm puts into it.

11. Where to Start

Make one command yours before you touch the other nine.

1. Choose the command your team will use this quarter, usually /ic-memo or /screen-deal.
2. Write down its last column, signed off by the partner who owns it.
3. Install financial-analysis and the plugin for a test group, and run the command on a redacted past deal.
4. Fork and rewrite that one skill, then test it against two past memos or screens.
5. Distribute it: your copy Required, Anthropic's original Not available, User-created skills off.
6. Set connectors to read-only and check the two Cowork approval settings.
7. Review usage analytics after 30 days, then pick the next command.

The Claude for private equity guide covers the plans and models around this work, and projects vs custom build covers what to do when skills reach their limits.