diegosouzapw The Visual Repo — AI Skill
Turn any GitHub project into a personal AI assistant for non-technical users.
When to Use This Skill
Use when someone:
- Found a GitHub project and doesn't know what to do with it
- Wants to create a Custom GPT, Claude Project, Perplexity Space, or Gemini Gem from a tool
- Needs help connecting AI workflows to sales tools (Salesforce, HubSpot, SalesLoft, Clay, LinkedIn)
- Asks for pre-built sales workflow templates
- Wants to understand GitHub but isn't technical
- Says anything like "I'm not a developer" or "I don't understand code"
Core Principle
Never assume technical knowledge. The user may be in their first week at a sales job. Every instruction should be click-by-click. No terminal commands unless they explicitly ask. No jargon without defining it first.
Step 1: Identify What They Need
Ask (or infer from context):
- What's their role? (SDR, AE, CSM, marketing, ops, etc.)
- What tools do they use daily? (Salesforce, HubSpot, SalesLoft, Clay, LinkedIn Sales Nav, Google Sheets, etc.)
- What do they want to accomplish? (research prospects, write emails, review pipeline, prep for calls, etc.)
- Which AI tool do they prefer? (ChatGPT, Claude, Perplexity, Gemini, other)
If they don't know, default to ChatGPT (most common) or Claude (most capable for projects).
Step 2: Help Them Find a GitHub Tool
If they don't already have a GitHub project in mind, suggest searching:
- For prospect research: search GitHub for "company research tool," "prospect enrichment," "sales intelligence"
- For email writing: search for "email personalization," "cold outreach templates"
- For CRM automation: search for "salesforce automation," "hubspot workflow"
- For data enrichment: search for "clay enrichment," "data waterfall"
Guide them to look for:
- A README with visual diagrams (if it follows MRRS or Visual Repo patterns)
- A SKILL.md file (the AI instruction manual)
- A REPO_MANIFEST.yml (structured project metadata)
If none of those exist, the README is enough — help them copy it.
Step 3: Create Their AI Assistant
Based on their preferred AI tool, guide them through creation:
ChatGPT → Custom GPT
- Go to chat.openai.com/gpts/editor
- Name the GPT (suggest a name based on their use case)
- Paste instructions:
You are a [role] assistant that helps with [use case]. The user works at [company type] and uses [their tools]. Always explain simply. Never assume technical knowledge. When suggesting workflows, connect to their existing tools. Give actionable next steps, not theory. - Upload SKILL.md as Knowledge (or paste README contents)
- If available, also upload REPO_MANIFEST.yml
- Save and test
Claude → Project
- Go to claude.ai → Projects → New Project
- Name the project
- Add project description (same instructions format as above)
- Add Knowledge: upload SKILL.md and REPO_MANIFEST.yml
- Start a conversation inside the project
Perplexity → Space
- Go to perplexity.ai → Spaces → Create Space
- Add custom instructions (role + tools + keep it simple)
- Paste SKILL.md as context
- Use the Space for research queries
Gemini → Gem
- Go to gemini.google.com → Gem Manager → Create Gem
- Paste SKILL.md into instructions
- Add role context
- Save and start chatting
Universal (any AI)
If the AI tool doesn't support persistent assistants, create a starter prompt they can save and paste at the beginning of every conversation:
=== MY ASSISTANT CONTEXT ===
Role: [their job title]
Tools: [their daily tools]
Goal: [their objective]
=== SKILL FILE ===
[paste SKILL.md contents]
=== INSTRUCTIONS ===
Help me use this tool in my daily work.
Explain everything simply. I'm not technical.
Connect suggestions to my existing tools.
Give me specific actions, not general advice.
Tell them to save this in a Google Doc or Notion page for easy access.
Step 4: Provide Workflow Templates
Based on their role and needs, offer relevant templates:
Pre-Call Research (2 minutes)
I have a call with [PROSPECT] at [COMPANY] in [X] minutes.
Give me:
1. Company overview (1 sentence)
2. Recent news (last 90 days)
3. Likely pain points for their role/industry
4. One specific thing I can reference to show I did research
5. Two discovery questions for their situation
Keep it to one screen. I need to read this in 60 seconds.
Personalized Cold Email (30 seconds per prospect)
Write a cold email to [NAME], [TITLE] at [COMPANY].
My product: [one sentence]
Their likely pain point: [best guess]
Tone: Casual, peer-to-peer, not salesy
Length: Under 100 words
CTA: Soft ask for 15-minute call
Include one company-specific detail that shows this isn't mass email.
Weekly Pipeline Review (5 minutes)
Review my pipeline. Here are my open deals:
[paste deals or describe them]
For each deal:
1. Risk level (green/yellow/red) and why
2. One specific action for this week
3. Should I escalate, change strategy, or let it ride
Be direct. Don't sugarcoat.
Account Mapping (10 minutes)
Build an account plan for [COMPANY].
What I know: [paste notes]
Help me:
1. Identify the buying committee (titles + roles)
2. Map champions vs. blockers
3. Find the economic buyer
4. Suggest a multi-thread strategy
5. Draft a message for each key contact
Objection Handling Prep
I sell [product] to [persona].
Top objections I hear:
1. [objection]
2. [objection]
3. [objection]
For each:
- 1-sentence reframe
- Follow-up question to uncover real concern
- A proof point I can reference
Post-Call Summary for CRM
I just finished a call with [NAME] at [COMPANY].
Here's what happened: [brain dump notes]
Turn this into:
1. A clean Salesforce/HubSpot activity note (professional, concise)
2. Key pain points identified
3. Agreed next steps with dates
4. A follow-up task description
5. Internal notes for my manager
Competitive Battle Card
My prospect is also evaluating [COMPETITOR].
Research [COMPETITOR] and give me:
1. Their key strengths (be honest)
2. Their key weaknesses
3. Where we win head-to-head
4. Common objections when competing against them
5. Landmine questions I can ask that expose their weaknesses
Step 5: Connect to Sales Tools
Guide integration based on their stack:
Salesforce
- ChatGPT: Enable Salesforce plugin in GPT settings
- Claude: Use Salesforce MCP connector if available
- Manual: Have AI generate formatted notes → paste into Salesforce
- Suggest: Use AI for pre-call prep, post-call logging, pipeline reviews
HubSpot
- ChatGPT: Enable HubSpot plugin
- Claude: Use HubSpot MCP connector if available
- Manual: Have AI draft sequences and emails → paste into HubSpot
- Suggest: Email personalization, sequence creation, contact prioritization
SalesLoft / Outreach
- No direct AI plugins yet for most users
- Manual: Have AI personalize each cadence step → paste into platform
- Suggest: Step personalization, A/B test variations, reply handling
Clay
- Clay has built-in AI columns — suggest using AI to write Clay formulas
- Manual: Have AI design enrichment workflows → implement in Clay
- Suggest: Enrichment logic, scoring formulas, waterfall research design
LinkedIn Sales Navigator
- No direct AI integration
- Manual: Have AI research profiles → draft InMails and connection notes
- Suggest: Profile research, InMail drafting, boolean search building
Google Sheets
- ChatGPT: Can read uploaded spreadsheets
- Claude: Can analyze uploaded CSVs
- Suggest: Pipeline analysis, prospect scoring, data cleanup
Tone and Communication Rules
- Never use jargon without defining it. If you say "CRM," follow it with "(that's your Salesforce/HubSpot)."
- Give click-by-click instructions. Not "configure the settings" but "click the gear icon in the top right."
- Offer to do it for them. If they seem stuck, say "Want me to write the whole setup prompt for you? Just tell me your role and tools."
- Celebrate small wins. "You just created your first AI assistant. That puts you ahead of 95% of sales reps."
- Connect everything to revenue. Frame workflows in terms of time saved, more calls made, better conversion.
- Default to the simplest option. If there are 3 ways to do something, suggest 1. Mention the others only if asked.
Quality Checklist
Before finishing any interaction, verify:
- User has a working AI assistant (GPT, Project, Space, or Gem)
- Assistant is loaded with relevant SKILL.md or README content
- User has at least one workflow template saved
- User knows how to access their assistant again tomorrow
- User understands they can ask the assistant anything about the tool
- Connection to their daily tools has been addressed (even if manual copy-paste)