Setup Guide

Using RoseMehl skills
requires Claude.ai.

Every RoseMehl AI agent is powered by Claude - Anthropic's AI. This guide covers how to set up your account, configure privacy settings, understand what Claude does and doesn't store, and get the most out of your AI agents.

⚠️
Claude.ai account required
RoseMehl skills are AI-powered prompts and systems built on top of Claude. To use them, you need a Claude.ai account at claude.ai. A free account works for most skills. Claude Pro ($20/month) is recommended for heavy usage, file uploads, and longer conversations. You do not share your Claude account with RoseMehl - your conversations stay in your own Claude workspace.
Step 1

Create your Claude.ai account

Go to claude.ai and sign up with your work email. The process takes under two minutes.

1
Visit claude.ai and click "Sign up"
Use your work email address - not a personal Gmail. This keeps your conversations clearly associated with your professional context and makes it easier to manage access if you add team members later.
2
Choose your plan
The Free plan works for getting started. If you're using RoseMehl skills daily or working with large documents (financial models, contracts, data room files), upgrade to Claude Pro at $20/month. Pro gives you significantly higher message limits, file upload capability (PDFs, Excel, Word), and priority access during peak hours.
3
Set up your profile name
Go to Settings → Profile and add your name and role. Claude uses this to personalise responses - a CFO gets different framing than a founder. Takes 30 seconds and meaningfully improves output quality.
4
Configure Privacy Settings
Go to Settings → Privacy. This is the most important step - covered in detail in Section 3 below.
FeatureFreePro ($20/mo)Team ($25/user/mo)
Basic conversations✓ Unlimited✓ Unlimited✓ Unlimited
File uploads (PDF, Excel, Word)Limited✓ Yes✓ Yes
Long context (200K tokens)No✓ Yes✓ Yes
Projects (saved context)No✓ Yes✓ Yes
Priority accessNo✓ Yes✓ Yes
Team collaborationNoNo✓ Yes
Admin controls + SSONoNo✓ Yes
Step 2 - Pro feature

Set up a Project for RoseMehl skills

Claude Projects (Pro and Team plans) let you create a persistent workspace with custom instructions and uploaded context. This is the single biggest productivity multiplier for using RoseMehl skills.

1
Create a new Project called "Finance Ops"
In the Claude sidebar, click Projects → New Project. Name it something like "Finance & Operations" or your company name. Every skill conversation you run inside this project will have access to your uploaded context.
2
Add your company context
Upload a short brief (1–2 pages) about your company: stage, industry, team size, revenue range, key systems (NetSuite, Salesforce, etc.), and your biggest operational challenges. Every skill conversation will now understand your specific context without you having to re-explain it every time.
3
Set Project Instructions
In Project Settings, add a custom instruction like: "You are working with [Company Name], a [stage] [industry] company. Our CFO is [name]. We use [systems]. Always respond with specific numbers and actionable recommendations." This transforms every skill from generic to specific.
4
Upload your key documents
Within the project, upload documents that skills will reference: your current financial model, org chart, chart of accounts, existing board pack template, or SOX control matrix. The FP&A, Close Automation, and Board Reporting skills immediately become more powerful when they can reference your actual documents.
Pro tip - one project per engagement type
  • Create a "Board Reporting" project with your board templates and prior packs uploaded
  • Create an "M&A Diligence" project with your deal framework and checklist
  • Create a "Legal Review" project with your standard contract templates
  • Each project remembers everything - you never re-upload the same context twice
Step 3 - Critical

Privacy settings you must configure

Before using any RoseMehl skills with real company data, configure these settings. Anthropic is transparent about how data is handled - here is what matters for professional and enterprise use.

Do this
Turn off conversation training
  • Go to Settings → Privacy
  • Turn off "Use my conversations to improve Claude"
  • This prevents your financial data from being used in Anthropic model training
  • Available on all plans including Free
  • Takes 5 seconds. Do it before your first skill run.
Understand this
What Anthropic stores
  • Conversations are retained for 30 days by default (adjustable)
  • Anthropic employees can access conversations for safety review
  • Data is stored in the US on AWS infrastructure
  • Free and Pro plans: data used for trust & safety review
  • Team and Enterprise: stronger data isolation
For enterprise use
Claude for Work (Team/Enterprise)
  • Conversations not used to train Anthropic models (default)
  • 30-day retention with admin control over deletion
  • SSO integration for access management
  • Audit logs for compliance requirements
  • BAA available for healthcare companies on Enterprise
  • SOC 2 Type II certified infrastructure
Never do this
What not to paste into Claude
  • Actual customer names, SSNs, or PII in bulk
  • Trade secrets or patentable IP descriptions
  • Passwords, API keys, or credentials
  • Unpublished M&A targets before announcement
  • Material non-public information (MNPI)
  • Health records (PHI) on non-BAA plans
For CFOs and finance teams - the right level of detail
  • Use revenue ranges ($X–$Y) rather than exact figures for highly sensitive numbers
  • Anonymise customer names in data you paste - "Customer A" not "Acme Corp"
  • For SOX and audit work, work from control descriptions not actual test results with findings
  • For M&A, use code names for targets until post-signing
  • Internal financial models are generally fine - they are not public information but are not trade secrets
Step 4

Getting the best results from RoseMehl skills

A RoseMehl skill is a precisely engineered prompt that turns Claude into a senior specialist. The output quality depends almost entirely on the specificity of what you give it.

1
Give context up front
Before running a skill, paste a 3–5 sentence summary of your situation. Example: "We are a 45-person Series B SaaS company at $8M ARR growing 3x. Our close takes 14 days. We use NetSuite, Salesforce, and Brex. The controller wants to get to 5 days." The skill will produce output calibrated to your specific situation, not a generic template.
2
Upload the relevant document
For model QA, upload your actual Excel model. For contract review, upload the actual contract. For board reporting, upload your prior board pack. Claude reads documents natively - you don't need to copy-paste. On Pro, you can upload PDFs, Excel files, Word documents, and CSVs.
3
Ask for specific formats
Tell Claude exactly what you want back. "Give me a table with three columns: task, automation classification, and estimated hours saved per month." Specific output format instructions are the single biggest driver of output quality.
4
Iterate - don't start over
Claude remembers the entire conversation. If the first output needs adjustment, say "The close savings number assumes a $150K salary - use $185K instead and recalculate." You are talking to a senior analyst who has all the context. Use it.
5
Save good conversations as templates
When a conversation produces exactly the right output, bookmark it. Next quarter, open that conversation, update the numbers, and run it again. The prompt, context, and format instructions are all preserved. Your best conversations become reusable workflows.
Reference

Anthropic's official privacy resources

RoseMehl does not control how Anthropic handles your data. Read the source directly:

↗
Anthropic Privacy Policy
Full legal privacy policy covering data collection, retention, processing, and your rights under GDPR, CCPA, and other privacy frameworks.
↗
Does Anthropic train on my conversations?
The definitive answer from Anthropic support - what the training opt-out covers, what it doesn't, and how it differs across plans.
↗
Anthropic Security page
SOC 2 Type II certification, infrastructure security, access controls, and enterprise security documentation.
↗
Your Claude Privacy Settings
Direct link to your personal privacy settings in Claude - turn off training data collection here.

Ready to use your first skill?

You have your Claude account configured. Explore our AI Agent Workforce and deploy your first specialist.

Explore AI Agents → Talk to Sophie first