graph LR
A["π¦ GitHub Project"] -->|"+ AI Tool"| B["π€ Your Personal<br/>AI Assistant"]
B --> C["π Connected to<br/>Your Sales Tools"]
style A fill:#a29bfe,color:#fff,stroke:none
style B fill:#00b894,color:#fff,stroke:none
style C fill:#0984e3,color:#fff,stroke:none
Find a tool. Create an AI assistant from it. Connect it to your workflow.
No coding. No terminal. Just your browser and an AI chat.
Someone sent you a GitHub link β or you found one yourself β and you want to actually use it. Not just read about it. Use it. Every day. Connected to the tools you already work in.
This page shows you how to go from "I don't understand GitHub" to "I have a custom AI assistant that researches prospects for me" β in about 15 minutes.
You only need one. Whichever you already use:
|
ChatGPT |
Claude |
Perplexity |
Gemini |
Other AI |
Here's everything you're about to do. It's simpler than it looks:
graph TD
A["1οΈβ£ FIND<br/>Find a useful project<br/>on GitHub"] --> B["2οΈβ£ GRAB<br/>Copy the SKILL.md<br/>or README"]
B --> C["3οΈβ£ CREATE<br/>Make a Custom GPT,<br/>Claude Project, or<br/>Perplexity Space"]
C --> D["4οΈβ£ TEACH<br/>Paste the skill into<br/>your AI assistant"]
D --> E["5οΈβ£ USE<br/>Talk to your assistant<br/>like a coworker"]
E --> F["6οΈβ£ CONNECT<br/>Link it to Salesforce,<br/>HubSpot, Clay, etc."]
style A fill:#a29bfe,color:#fff,stroke:none
style B fill:#6c5ce7,color:#fff,stroke:none
style C fill:#fd79a8,color:#fff,stroke:none
style D fill:#e17055,color:#fff,stroke:none
style E fill:#00b894,color:#fff,stroke:none
style F fill:#0984e3,color:#fff,stroke:none
A website where people share free tools. That's it. Like an app store, except everything is free and you can see how it's built.
The page you see when you open a project. It's the instructions. You're reading one right now.
An instruction manual written for AI. When you give this file to ChatGPT or Claude, your AI instantly becomes an expert on that tool.
A structured summary of the entire project. Think of it as the nutrition label. Hand it to any AI tool and ask "what does this project do?" β you'll get a perfect answer every time.
| Word | What It Means | Real-World Equivalent |
|---|---|---|
| Repo | A project folder | A folder on your computer |
| README | The welcome page | The back of a cereal box |
| Clone | Download a copy | Saving a Google Doc to your computer |
| Commit | Save your changes | Hitting "Save" in Word |
| Branch | A separate copy to experiment with | A rough draft |
Want to see if you get it? Try this right now:
- Go to github.com/MahoneyContextProtocol/claude-cleanup
- Read the first 3 paragraphs of the README
- Look at the colorful diagram at the top
- You now understand that project. See? Not scary.
This is the main event. Pick your AI tool and follow the steps.
A Custom GPT is a version of ChatGPT that's pre-loaded with specific knowledge. Once you create it, you can talk to it forever β it already knows the tool.
-
Go to chat.openai.com/gpts/editor (you need ChatGPT Plus)
-
Name it something you'll remember
Example: "Prospect Research Assistant"
-
In the Instructions box, paste this starter prompt:
You are my personal assistant for [describe your use case]. I'm going to give you a skill file from a GitHub project. Learn it completely. Then help me use it in my daily work. My role: [your job title] Tools I use: [Salesforce / HubSpot / SalesLoft / Clay / etc.] My goal: [what you want to accomplish] Always explain things simply. I'm not technical. When suggesting workflows, connect them to my existing tools. -
In the Knowledge section, upload the
SKILL.mdfile from the GitHub project- Go to the GitHub project
- Click on
SKILL.md - Click the download button (β icon) or copy all the text
- Upload it to your GPT's Knowledge section
-
Click Save β Choose "Only me" or "Anyone with a link"
-
Start chatting. Your GPT now knows the tool. Ask it anything:
"How do I use this to research prospects before cold calls?" "Build me a morning routine using this tool and Salesforce" "Connect this to my HubSpot workflow"
graph LR
A["π Copy SKILL.md<br/>from GitHub"] --> B["βοΈ Create GPT<br/>at chat.openai.com"]
B --> C["π Upload SKILL.md<br/>as Knowledge"]
C --> D["π¬ Chat with your<br/>new expert assistant"]
style A fill:#6c5ce7,color:#fff,stroke:none
style B fill:#a29bfe,color:#fff,stroke:none
style C fill:#fd79a8,color:#fff,stroke:none
style D fill:#00b894,color:#fff,stroke:none
A Claude Project is a workspace where Claude remembers context between conversations. It's perfect for building persistent assistants.
-
Go to claude.ai β Click Projects in the sidebar β New Project
-
Name it
Example: "Sales Research Assistant"
-
Add a Project Description (this tells Claude its role):
You are my personal assistant for sales prospecting and research. My role: [your job title, e.g., "SDR at a SaaS company"] Tools I use daily: [Salesforce / HubSpot / SalesLoft / Clay / LinkedIn Sales Nav] My goal: [e.g., "Research prospects quickly before cold calls"] Always explain things simply. I'm not technical. When suggesting workflows, connect them to my existing tools. Suggest specific actions I can take right now. -
Add Knowledge β Click "Add content" β Upload or paste the
SKILL.mdfile from the GitHub project -
Also add the
REPO_MANIFEST.ymlif the project has one (this gives Claude structured data about the tool) -
Start a conversation inside the project. Claude now knows the tool permanently:
"What does this tool do and how can I use it for prospect research?" "Build me a workflow that combines this with my Salesforce process" "I have a call with [Company] in 10 minutes β help me prep"
graph LR
A["π Copy SKILL.md<br/>from GitHub"] --> B["π Create Project<br/>at claude.ai"]
B --> C["π Add SKILL.md<br/>to Project Knowledge"]
C --> D["π¬ Every conversation<br/>in this project<br/>has full context"]
style A fill:#6c5ce7,color:#fff,stroke:none
style B fill:#a29bfe,color:#fff,stroke:none
style C fill:#fd79a8,color:#fff,stroke:none
style D fill:#00b894,color:#fff,stroke:none
Perplexity Spaces let you create focused research areas with custom instructions and sources.
-
Go to perplexity.ai β Spaces β Create Space
-
Name it and add a description
Example: "Prospect Research" / "Pre-call research using AI tools"
-
In the Custom Instructions, paste:
I'm a [job title] at [company]. I use [your tools: Salesforce, HubSpot, etc.]. Help me use GitHub tools for my sales workflow. Always explain simply β I'm not technical. Give me actionable steps connected to my existing tools. -
Add the
SKILL.mdas a source β paste the text into the Space's context or upload the file -
Search and chat inside the Space. Perplexity will combine its web search with your tool knowledge:
"Research [Company Name] and tell me what I should know before my call, using the approach from my skill file"
Google Gems are custom Gemini assistants with persistent instructions.
-
Go to gemini.google.com β Gem Manager β Create Gem
-
Paste the SKILL.md contents into the Gem's instructions
-
Add your role context:
"I'm a [title]. Help me use this tool with [my existing tools]. Keep it simple."
-
Save and start chatting with your Gem
If your AI tool doesn't have custom assistants, you can still use the Persistent Prompt Method:
-
Save this text somewhere you can copy it quickly (Google Doc, Notion, Notes app):
=== MY AI ASSISTANT CONTEXT === My role: [your job title] My tools: [list your daily tools] My goal: [what you want help with] Below is a skill file from a GitHub tool I want to use. Help me apply it to my daily workflow. Always explain simply β I'm not technical. === SKILL FILE === [paste SKILL.md contents here] -
Every time you start a new chat, paste this at the beginning
-
Then ask your question
It's not as clean as a Custom GPT or Claude Project, but it works with any AI tool β Microsoft Copilot, Mistral, Grok, you name it.
This is where it gets powerful.
graph TD
AI["π€ Your AI Assistant<br/>(GPT / Claude Project / Gem)"] --> SF["Salesforce<br/><i>Update records, log calls,<br/>pull account data</i>"]
AI --> HS["HubSpot<br/><i>Contact enrichment,<br/>sequence triggers</i>"]
AI --> SL["SalesLoft / Outreach<br/><i>Personalize cadences,<br/>research before steps</i>"]
AI --> CL["Clay<br/><i>Prospect enrichment,<br/>waterfall research</i>"]
AI --> LI["LinkedIn Sales Nav<br/><i>Profile research,<br/>connection notes</i>"]
AI --> SH["Google Sheets<br/><i>Pipeline tracking,<br/>research databases</i>"]
style AI fill:#6c5ce7,color:#fff,stroke:none
style SF fill:#0984e3,color:#fff,stroke:none
style HS fill:#e17055,color:#fff,stroke:none
style SL fill:#00b894,color:#fff,stroke:none
style CL fill:#fd79a8,color:#fff,stroke:none
style LI fill:#0984e3,color:#fff,stroke:none
style SH fill:#00b894,color:#fff,stroke:none
You don't connect tools by writing code. You connect them by describing workflows to your AI assistant. The AI figures out the steps.
Ask your AI assistant:
"I just got off a call with [prospect]. Help me write a Salesforce activity note that includes: key pain points discussed, next steps, and a follow-up task for next Tuesday."
"Pull up what we know about [Company] from our Salesforce records and combine it with what you can research online. I have a call in 15 minutes."
For ChatGPT with Salesforce plugin: Enable the Salesforce plugin in your Custom GPT settings. Your GPT can then read and write Salesforce data directly.
For Claude with MCP: If your company uses Claude with Salesforce MCP connector, your Claude Project can access Salesforce data directly.
"I need to send a follow-up sequence to the 12 leads from last week's webinar. Draft 3 personalized email variations I can paste into HubSpot sequences."
"Look at my HubSpot contact list for [segment] and help me prioritize who to call first based on engagement signals."
"I'm on step 3 of my SalesLoft cadence for [prospect]. Research their company and help me personalize this touchpoint."
"Write me 5 variations of a cold email for [persona] that I can A/B test in Outreach."
"I have a Clay table of 200 prospects. Help me write a formula column that categorizes them by likely pain point based on company size and industry."
"Build me a Clay enrichment workflow that takes a company name and returns: recent funding, tech stack, and hiring signals."
"I'm about to send an InMail to [prospect]. Research their profile, recent posts, and company news. Help me write a message that references something specific."
"I found 30 results in Sales Nav for [search]. Help me narrow to the 10 most likely to respond based on their activity signals."
Copy any of these prompts into your AI assistant to get started immediately:
π₯ Pre-Call Research (2 minutes)
Paste into your Custom GPT / Claude Project:
I have a call with [PROSPECT NAME] at [COMPANY] in [X] minutes.
Research them and give me:
1. Company overview (1 sentence)
2. Recent news (last 90 days)
3. Their likely pain points based on their role and industry
4. One specific thing I can reference to show I did my homework
5. Two discovery questions tailored to their situation
Keep it to one screen. I need to read this in 60 seconds.
π§ Personalized Cold Email (30 seconds per prospect)
Write a cold email to [NAME], [TITLE] at [COMPANY].
About my product: [one sentence about what you sell]
Their likely pain point: [your best guess]
Tone: Casual, peer-to-peer, not salesy
Length: Under 100 words
CTA: Soft ask for a 15-minute call
Include one specific detail about their company or role
that shows this isn't a mass email.
π Weekly Pipeline Review (5 minutes)
Help me review my pipeline. Here are my open deals:
[Paste your pipeline data or describe your deals]
For each deal, tell me:
1. Risk level (green/yellow/red) and why
2. One specific next action I should take this week
3. If I should escalate, change strategy, or let it ride
Be direct. Don't sugarcoat.
π― Account Mapping (10 minutes)
I'm building an account plan for [COMPANY].
What I know so far: [paste any notes]
Help me:
1. Identify the likely buying committee (titles and roles)
2. Map potential champions vs. blockers
3. Find the economic buyer
4. Suggest a multi-thread strategy
5. Draft a personalized message for each key contact
π Objection Handling Prep
I sell [your product/service] to [your buyer persona].
The top 5 objections I hear are:
1. [objection]
2. [objection]
3. [objection]
4. [objection]
5. [objection]
For each one, give me:
- A 1-sentence reframe
- A follow-up question that uncovers the real concern
- A customer proof point I can reference (make it realistic)
π Post-Call CRM Summary
I just finished a call with [NAME] at [COMPANY].
Here's what happened: [brain dump your 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
βοΈ 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
You want to make your own projects this accessible. Here's the standard.
Everything you've seen on this page follows a pattern called MRRS (Machine-Readable Repo Standard). It's a set of templates that makes any GitHub project understandable by everyone.
Full standard + templates: github.com/MahoneyContextProtocol/mrrs
A SKILL.md file turns any project into something AI tools can instantly understand. The format works with: ChatGPT, Claude, Perplexity, Gemini, Cursor, Windsurf, Blackbox AI, Cline, Aider, Continue.dev, OpenCode, Mistral, and any MCP-compatible client.
graph TD
P1["π¨ Visual First<br/><i>Diagrams before text</i>"] --> P2["π Levels, Not Walls<br/><i>Start simple, go deep</i>"]
P2 --> P3["πͺ Every Path Leads Somewhere<br/><i>Non-tech, dev, AI β all welcome</i>"]
P3 --> P4["π€ AI-Native<br/><i>SKILL.md + MANIFEST.yml<br/>for instant machine parsing</i>"]
P4 --> P5["π Tool-Connected<br/><i>Workflows link to<br/>real tools people use daily</i>"]
style P1 fill:#fd79a8,color:#fff,stroke:none
style P2 fill:#a29bfe,color:#fff,stroke:none
style P3 fill:#00b894,color:#fff,stroke:none
style P4 fill:#0984e3,color:#fff,stroke:none
style P5 fill:#6c5ce7,color:#fff,stroke:none
git clone https://github.com/MahoneyContextProtocol/mrrs.git /tmp/mrrs
cp -r /tmp/mrrs/template/* your-project/
cp -r /tmp/mrrs/template/.github your-project/
# Fill in {placeholders}, commit, pushOr point any AI coding assistant at the MRRS SKILL.md and say: "Apply MRRS to this repo."
graph TD
subgraph "π FIND"
F1["Search GitHub for<br/>a tool you need"]
end
subgraph "π GRAB"
G1["Find SKILL.md<br/>or README"]
end
subgraph "π€ CREATE"
C1["Make a Custom GPT<br/>or Claude Project<br/>or Perplexity Space"]
end
subgraph "π TEACH"
T1["Upload SKILL.md<br/>to your AI assistant"]
end
subgraph "π¬ USE"
U1["Chat with your<br/>assistant daily"]
end
subgraph "π CONNECT"
X1["Link to Salesforce,<br/>HubSpot, Clay, etc."]
end
F1 --> G1 --> C1 --> T1 --> U1 --> X1
style F1 fill:#a29bfe,color:#fff,stroke:none
style G1 fill:#6c5ce7,color:#fff,stroke:none
style C1 fill:#fd79a8,color:#fff,stroke:none
style T1 fill:#e17055,color:#fff,stroke:none
style U1 fill:#00b894,color:#fff,stroke:none
style X1 fill:#0984e3,color:#fff,stroke:none
MIT β Free to use, modify, and share.
Created by Joe Mahoney at Cloud Latitude
β Star this repo if it made GitHub less scary.
π Share it with someone who thinks GitHub isn't for them.