Every financial firm has its own way of doing things — its own models, templates, terminology, and deal processes. These plugins are designed as starting points that you customize to match how your team actually works. This page walks you through every type of customization with detailed examples.The key insight is that everything is file-based — markdown and JSON. There is no code to write, no infrastructure to deploy, no build steps to run. If you can edit a text file, you can customize these plugins.
Out of the box, the plugins follow industry-standard practices. But “industry standard” is not the same as “how we do it at Goldman” or “how we do it at KKR.” Here are real examples of firm-specific differences:
Formatting: Some firms use Times New Roman; others use Calibri. Some report EBITDA excluding stock-based compensation; others include it.
Terminology: At one firm, “the book” means the CIM; at another, “the model” always refers to the LBO model.
Process: Some firms require 3 IC rounds; others require 2. Some use fully competitive auctions; others prefer negotiated deals.
Data sources: Your firm might use Bloomberg instead of FactSet, or have an internal data warehouse that is the authoritative source for certain metrics.
When you encode these preferences into the plugin files, Claude stops generating generic output and starts producing work that fits your firm’s standards.
Edit .mcp.json to point Claude at your specific data providers and internal tools. The financial-analysis plugin ships with 11 pre-configured MCP integrations, but you can replace, add, or remove any of them.Default .mcp.json structure:
Only configure the providers you have active subscriptions for. Claude will use whatever sources are available and will not attempt to connect to unconfigured endpoints.
This is where customization has the biggest impact. Drop your firm’s terminology, deal processes, and formatting standards directly into skill files. Claude will draw on this context automatically when relevant.
Skill files are located at skills/[skill-name]/SKILL.md within each plugin directory. Open any skill file in a text editor and add a “Firm Context” section:
## Firm Context### Our Formatting Standards- All models use Times New Roman 11pt, blue inputs, black formulas- EBITDA is always reported excluding stock-based compensation- Our deal team uses "Project [Name]" for all deals -- never the company name in external materials### Our Deal Process- LOI typically signed within 2 weeks of initial management meeting- IC process requires 3 rounds: screening, preliminary IC, and final IC- All comps analyses must include our standard footnote: "Source: [Provider], as of [Date]"### Internal Terminology- "The book" = CIM (Confidential Information Memorandum)- "The model" always refers to the LBO model, not a DCF- "Red process" = fully competitive auction with process letter
In financial-analysis/skills/comps-analysis/SKILL.md, you might add:
## Firm-Specific Peer Selection RulesWhen selecting comparable companies for our analyses:1. Always include [Firm Name]'s core coverage universe first2. Exclude companies with market cap < $500M unless specifically requested3. For healthcare comps, always check with our sector team before finalizing the peer group4. All comps must include NTM (next twelve months) estimates, not just LTM## Our Statistical Benchmarks- We report Max, 75th, Median, 25th, Min for all metrics- Never report simple averages -- always use median- Flag any outlier (>2 standard deviations) with an asterisk
## Firm-Specific LBO Conventions### Debt Structure Defaults- Senior Secured: 4.0x EBITDA at SOFR + 350bps- Subordinated: 1.5x EBITDA at 8.5% fixed- We never model mezzanine above 2.0x EBITDA without explicit approval### Returns Thresholds- Minimum target IRR: 20% gross- Target MOIC: 2.5x or higher- We flag any deal where > 50% of returns come from multiple expansion### Presentation Conventions- Always show both gross and net IRR- Sensitivity tables: 7x7 grid with entry multiple vs. exit multiple- Include management incentive plan (MIP) dilution in all returns
Use /ppt-template to teach Claude your firm’s branded PowerPoint layouts. Once created, the generated skill captures your template’s exact slide dimensions, placeholder positions, and layout structure.
1
Provide Your Template File
Run the command with your PowerPoint template:
/ppt-template path/to/your-firm-template.pptx
2
Claude Analyzes the Template
Claude extracts layout indices, placeholder positions (exact x/y coordinates), content area boundaries, and slide dimensions.
3
A New Skill is Generated
A self-contained skill is created:
your-firm-ppt-template/├── SKILL.md # Full instructions with exact coordinates└── assets/ └── template.pptx # Your template file bundled with the skill
4
Every Deck Matches Your Style Guide
Going forward, any presentation Claude creates will use your firm’s exact layouts, fonts, colors, and placeholder positions automatically.
In financial-analysis/skills/dcf-model/SKILL.md, you could add:
## Firm-Specific DCF Conventions### WACC Calculation- We always use the 10-year Treasury as the risk-free rate- Equity risk premium: use Duff & Phelps Supply-Side ERP (currently 5.5%)- Size premium: apply Duff & Phelps size premium based on market cap decile- Beta: use 2-year weekly raw beta, not 5-year monthly### Terminal Value- We prefer the exit multiple method over perpetuity growth- Default exit multiple: median of peer group EV/EBITDA (from our comps)- Never use a terminal growth rate above 3.0% without explicit justification### Sensitivity Analysis- Primary table: WACC vs. Exit Multiple (always)- Secondary table: Revenue Growth vs. EBITDA Margin (always)- Third table: only if specifically requested
{ "name": "your-plugin-name", "version": "0.1.0", "description": "Description of what this plugin does", "author": { "name": "Your Firm" }}
Command file example (commands/your-command.md):
---description: One-line description of what this command doesargument-hint: "[argument description]"---# Command NameBrief description of the command.## Workflow### Step 1: [First step]...### Step 2: [Second step]...
Skill file example (skills/your-skill/SKILL.md):
---name: your-skill-namedescription: When to use this skill -- trigger conditions and use cases.---# Skill Name## OverviewWhat this skill does and when it applies.## Workflow### Step 1: [First step][Instructions for Claude]### Step 2: [Second step][Instructions for Claude]## Quality Checklist- [ ] Check 1- [ ] Check 2
When adding new skills or commands, include clear trigger conditions in the skill description so Claude knows when to apply them automatically. The description field is what Claude uses to decide whether a skill is relevant to a given task.
Create .local.md files for user-specific configuration that should not be committed to a shared repository:
financial-analysis/└── .claude/ └── user.local.md # Personal configuration (gitignored)
user.local.md example:
# My Personal Configuration## My Coverage UniverseI cover enterprise software companies. When building comps, default to:- Salesforce (CRM), ServiceNow (NOW), Workday (WDAY), HubSpot (HUBS)## My Data AccessI have subscriptions to: FactSet, PitchBook, AieraI do NOT have access to: Daloopa, Chronograph## My Preferred Output Formats- Always use Excel (.xlsx) for models, never CSV- PowerPoint for presentations, never Google Slides- All currency in USD unless I specify otherwise
claude plugin install financial-analysis@financial-services-plugins
2
Locate Plugin Files
Find the plugin directory in your local Claude setup. The exact location depends on your installation, but you can search for the .claude-plugin directory.
3
Edit .mcp.json
Replace API key placeholders with your actual credentials. Remove providers you do not use. Add internal data sources.
4
Edit Skill Files
Open any SKILL.md file and add your firm context sections. Focus on the skills you use most often first (typically comps, DCF, and LBO).
5
Create user.local.md
Add your personal preferences in .claude/user.local.md. This file is gitignored so it stays local to your machine.
6
Test Your Customizations
Start a new session and run a workflow to verify your customizations are being picked up. For example, run /comps AAPL and check that your firm-specific peer selection rules are being followed.