Transform Claude into your personal growth hacker. Copy, paste, and deploy these 10 AI workflows to scrape, engage, and convert high-ticket LinkedIn prospects on autopilot.
Part 1: Laying the Foundation: ICP Synthesis & Profile Optimization (Skills 1-2)
Most founders treat Claude like a search engine.
They type a generic question. They get a generic answer. They stay broke.
If you want to automate LinkedIn lead generation, you need to build a machine.
That starts with treating Claude like a junior employee.
We do not want random outputs. We want a highly constrained, hyper-specific brain trained on your exact business.
This means using Claude Projects and Artifacts.
The 5-Point Artifact Setup
Before you execute a single workflow, you need guardrails.
Open a new Claude Project. You need to configure the system instructions.
If you skip this, Claude will hallucinate. Use this exact 5-point structure:
→ Role: "You are a top-tier B2B growth hacker specializing in LinkedIn."
→ Context: "We sell custom AI agents to SaaS founders."
→ Tone: "Blunt, authoritative, zero corporate fluff. Talk like an operator."
→ Output Format: "Staccato bullet points. No paragraphs over 2 lines."
→ Constraints: "Never use the words alignment, innovative, or comprehensive."
This setup kills bad habits before they happen.
Now, we feed the machine.
Skill 1: The Raw ICP Data Synthesizer
Defining your Ideal Customer Profile (ICP) usually takes weeks of manual labor.
Consultants charge $10,000 for it. We are going to automate it in 45 seconds.
The mistake everyone makes? Feeding Claude generic market research.
Garbage in, garbage out.
If you ask Claude to "research SaaS founders," it spits out generic marketer-speak.
You need raw, bleeding-neck voice-of-customer data.
Dump your Gong transcripts, Zendesk tickets, and raw discovery calls directly into the Project knowledge base.
Let's say you are an AI builder.
You upload five raw discovery call transcripts with SaaS founders.
You trigger Claude to hunt for exact phrases and complaints.
It doesn't just tell you they want "better efficiency."
It extracts the exact quote: "My AWS API costs are bleeding us dry and my devs are bottlenecked trying to integrate OpenAI."
You now own their exact vocabulary.
Skill 2: The High-Converting Profile Copywriter
Your LinkedIn profile is not a resume.
It is a high-converting landing page.
If it reads like a CV, high-ticket prospects will bounce immediately.
We use the exact ICP data extracted in Skill 1 to rewrite your entire profile.
We delegate the heavy lifting to Claude.
We force it to map the bleeding-neck pain points directly to your headline and About section.
The old way: "Experienced software engineer passionate about AI solutions."
The new way: "I audit your SaaS architecture and kill runaway API costs using custom AI agents."
One gets ignored. The other books meetings.
Troubleshooting Your Output
If Claude's output sounds like a corporate brochure, you failed the data step.
Never ask Claude to guess what your market wants.
Force it to read the transcripts.
Raw data is the only currency that matters in prompt engineering.
Deliverable: Claude Project Setup Checklist
Follow these steps to build your foundation:
→ Create a new Claude Project named "LinkedIn Lead Machine"
→ Paste the 5-Point Artifact Setup into the Custom Instructions
→ Upload 3-5 raw discovery call transcripts (TXT or PDF)
→ Upload your current LinkedIn profile text
→ Run the two Prompt Templates below sequentially
The ICP Synthesizer Prompt Template
Copy and paste this into your configured Project:
Analyze the attached raw call transcripts. Act as a forensic data analyst. Extract the top 3 most painful bottlenecks these prospects face. Give me the exact quotes they use when complaining about these problems. Then, generate a 1-page ICP cheat sheet formatting the data into: Pain Points, False Solutions They Tried, and Trigger Events that make them buy. Output this as an Artifact.
The Profile Conversion Prompt Template
Once the ICP Artifact is generated, run this prompt:
Take the ICP cheat sheet we just built. Act as an elite direct-response copywriter. Rewrite my attached LinkedIn profile. Turn it into a landing page that speaks directly to these exact pain points. Give me 3 options for a punchy headline. Rewrite the About section using short, staccato sentences. Focus purely on the ROI I deliver to the prospect. Output as an Artifact.
Quick-Start Tip For Skimmers
Want results in the next 5 minutes? Do this right now.
Open Claude. Create a Project. Drop in 3 recent sales calls.
Run the ICP Synthesizer Prompt.
You will instantly possess a 1-page prospect cheat sheet that is more accurate than a 50-page marketing agency report.
You just eliminated 40 hours of manual labor.
Your foundation is set. Next, we build the engine.
Before You Continue: Why We Verify Subscribers
Quick note.
We have had multiple instances of bots scraping this content and people republishing it without authorization.
This guide took dozens of hours to build. It is free. But it is not public domain.
To keep the full playbook available for real humans (and out of the hands of content thieves), we require a valid email address to unlock the remaining sections.
It takes 3 seconds. No spam. No upsells. Just proof you are a real person.
Enter your email below to unlock the full guide.
Most LinkedIn content is garbage.
It reads like a press release written by a nervous intern.
If you use Claude out of the box, you will sound like everyone else.
We don't do that. We use Claude 3.5 Sonnet to build an Authority Engine.
This engine crushes writer's block.
It generates high-resonance, non-robotic thought leadership that actually drives inbound.
Skill 3: The Hook-First Post Generator
Virality leaves clues. You do not need to invent new formats.
You just need to reverse-engineer what already works.
Most founders treat AI like a slot machine. They type: "Write a post about AI."
The result is a wall of text that gets zero impressions.
Instead, we force Claude to dissect the pacing, structure, and hook of top creators.
Example: The Solo Founder Automation Post
Stop telling people "AI saves time." Show them the blood and guts of the build.
Let's say you built a Make.com + OpenAI automation to score leads.
Don't write a generic summary about efficiency.
Force Claude to break it down into a digestible step-by-step carousel.
We move the reader from manual data entry (the Old Way) to complete delegation (the New Way).
The PAS-O Post Prompt
Copy and paste this exact prompt into Claude.
Problem grabs attention. Agitation twists the knife. Solution provides the build. Outcome proves the ROI.
Context: I am a solo founder building AI systems.
Task: Write a LinkedIn post using the PAS-O framework (Problem, Agitation, Solution, Outcome).
Topic: Building a Make.com + OpenAI lead scoring automation.
Constraints:
- Use short sentences. Maximum 2 lines per paragraph.
- Start with a punchy hook that agitates a specific pain point.
- Give away the exact technical steps for free.
- End with a measurable outcome.
- Ban these words: dig, mosaic, alignment, vital, silver bullet.
Audit the output. Tweak the variables. Post.
Skill 4: The Automated Engagement Copilot
Posting is only half the battle.
To trigger inbound, you must hijack attention from accounts larger than yours.
You do this by leaving top-1% comments on massive creator posts.
Commenting "Great post!" or "Thanks for sharing!" kills your credibility.
You need comments that siphon their audience directly to your profile.
The 3-C Commenting Framework
We use Claude to draft comments based on three pillars: Compliment, Challenge, Contribute.
Never comment on a big account without running this framework first.
Context: I want to leave a high-value comment on a massive LinkedIn post.
Task: Read the attached post. Write 3 comment variations using the 3-C Framework.
Variation 1 (Compliment): Validate their premise with a quick personal anecdote.
Variation 2 (Challenge): Respectfully disagree with a minor point and offer an alternative angle.
Variation 3 (Contribute): Add one highly technical, actionable step they missed.
Constraints: Keep it under 4 sentences. Sound conversational. No corporate fluff.
Troubleshooting: The AI Vocabulary Trap
If you forget this step, your content will fail.
LLMs naturally gravitate toward bloated, academic language. They revert to the average.
If you do not explicitly ban words, Claude will ruin your post.
Always include a negative constraint list in your prompts.
Ban words like "nurture," "boost," "environment," and "release."
If a word sounds like it belongs in a Fortune 500 mission statement, kill it.
Quick-Start Tip for Skimmers
Want to skip the prompt engineering entirely?
Find a viral post from a competitor in your feed.
Paste it into Claude with this exact command:
"Reverse-engineer the hook structure, rhythm, and pacing of this post. Now, apply that exact mathematical structure to [Your Topic]."
Deliverable: 30-Day Content Prompt Matrix & Commenting Framework
Stop guessing what to post. Automate the ideation phase.
Feed this matrix into Claude on Sunday to build your entire weekly backlog.
→ Monday (Tear-Down): "Analyze this failed outbound campaign. Write a post on why it failed."
→ Tuesday (Technical Build): "Draft a step-by-step guide on my latest Make.com workflow."
→ Wednesday (Contrarian Take): "Take the popular opinion on [Industry Topic] and argue the exact opposite."
→ Thursday (Case Study): "Convert this client win into a PAS-O framework post."
→ Friday (Resource Drop): "List 5 Claude skills I use daily. Format as a bulleted cheat sheet."
Run these prompts. Audit the output.
You just replaced your social media manager.
Part 3: Laser-Targeted Prospecting: Search Logic & Profile Extraction (Skills 5-6)
Most founders treat LinkedIn Sales Navigator like a slot machine.
They type "Founder" into the search bar, hit enter, and pray.
That is manual labor. It is slow. It burns hours.
We are building an automated machine.
Machines require strict, mathematically precise logic.
You need to move from manual scrolling to programmatic prospecting.
We do this by forcing Claude to write advanced search parameters.
Then, we use Claude to parse the messy data we scrape.
Skill 5: The Advanced Sales Navigator Boolean Builder
Basic searches yield basic prospects.
If you search "CEO," you get the exact same list as your competitors.
You are swimming in a red ocean of spam.
You need to hunt for hidden intent.
You do this by combining obscure keywords with strict exclusions.
Let's say you sell automation services or custom AI builds.
You don't want a generic VP of Operations.
You want the VP of Ops who is actively drowning in manual work.
You want the operator explicitly mentioning "Zapier," "Airtable," or "process bottleneck."
Writing these massive Boolean strings manually causes headaches.
Claude builds complex, error-free Boolean strings in seconds.
You just need to feed it the exact parameters.
Copy and paste this prompt to build your search logic.
The Boolean Builder Prompt
I need to build an advanced LinkedIn Sales Navigator Boolean search string.
My target persona is: [Target Title, e.g., VP of Operations, COO, Head of Ops]
They must include at least one of these exact keywords in their profile:
- [Keyword 1, e.g., Zapier]
- [Keyword 2, e.g., Airtable]
- [Keyword 3, e.g., process bottleneck]
They must NOT include these keywords (dealbreakers):
- [Exclusion 1, e.g., Consultant]
- [Exclusion 2, e.g., Fractional]
- [Exclusion 3, e.g., Agency]
Generate 5 different variations of a strict Boolean string I can paste directly into Sales Navigator.
Start broad, then make each variation progressively more niche.
Format them clearly. Do not explain the logic. Just output the raw strings.
Claude will generate strings like: ("VP of Operations" OR "COO") AND ("Zapier" OR "Airtable") NOT ("Consultant" OR "Fractional").
Paste that into LinkedIn. Watch your prospect quality spike instantly.
Skill 6: The Unstructured Profile Data Extractor
You have your hyper-targeted prospect list.
Now you need to extract their profile data to trigger your outreach.
But LinkedIn profile data is incredibly messy.
It is massive walls of unstructured text, emojis, and weird formatting.
If you feed raw profile text into a Make.com or Zapier webhook, it breaks.
You cannot automate cold email sequences with raw text.
You need structured data. You need JSON formatting.
Claude is a master at parsing garbage text into clean, mapped fields.
You scrape the raw profile text using a tool like Phantombuster or Apify.
You pass that raw text to Claude via API.
Claude extracts exactly what you need and outputs pure JSON.
This allows you to pipe hyper-personalized variables straight into Instantly or Smartlead.
The Profile Extraction JSON Schema Prompt
You are an expert data extraction agent.
I will provide raw text scraped from a LinkedIn profile.
Your job is to extract specific data points and format them strictly as a JSON object.
Do not output any conversational text. Only output the JSON.
Extract these exact fields:
- "FirstName": (String - extract just the first name)
- "CurrentRole": (String - their exact current job title)
- "CorePainPoint": (Analyze their "About" section. What operational problem are they trying to solve? Max 5 words.)
- "RecentPostTopic": (Analyze their recent activity. What specific topic did they post about recently? Max 3 words.)
Here is the raw profile text:
[INSERT RAW TEXT]
Troubleshooting Your Prospecting Engine
This system is lethal. It kills manual research.
But automation operators make two fatal errors when building this.
First: Using Boolean variables that are too broad.
If your Claude-generated string just says "Founder" OR "CEO", you failed.
→ Always force Claude to include specific niche software tools.
→ Always force Claude to exclude your competitors and consultants.
→ Always test the string in Sales Nav. If the list is junk, adjust the prompt.
Second: Failing to enforce strict JSON formatting in Claude.
LLMs are conversational by nature. They love to chat.
They want to say, "Here is the JSON data you requested!"
That single sentence will instantly crash your Make.com scenario.
Your downstream automations rely on perfect machine syntax.
→ Use the system prompt to completely ban conversational text.
→ Add this command: "Return ONLY valid JSON. No markdown formatting. No backticks."
The Quick-Start Blueprint
Skip the theory. Execute this workflow right now.
→ Open Claude.
→ Feed it your ideal prospect title.
→ Give it 3 highly niche software tools your ICP uses.
→ Ask for 5 Sales Navigator Boolean strings.
Grab the tightest string. Run the search.
You just delegated 10 hours of manual prospecting to a machine.
Part 4: Inbound-Level Outbound: Connection Requests & DM Sequences (Skills 7-8)
LinkedIn inboxes are a warzone.
Every day, your prospects get hammered by automated pitches.
"I noticed we are both in the AI space."
Delete.
"Would love to explore partnerships."
Block.
If you sound like a bot, you die like a bot.
We need outreach that feels 1-to-1. Hand-crafted. Personal.
But we refuse to do the manual labor.
Enter Claude.
The Inbox Reality
Standard cold outreach fails because it is lazy.
It relies on volume to compensate for a lack of relevance.
We flip the script.
We use AI to achieve inbound-level relevance on outbound campaigns.
We extract the data. We feed it to Claude.
Claude writes a message so specific it is impossible to ignore.
Skill 7: The Non-Spam Connection Request Writer
You have 300 characters to prove you are human.
Do not pitch here. Ever.
Your only goal is a click on the "Accept" button.
We optimize for a >40% acceptance rate using psychological triggers.
We use a framework I call the 300-Character Trojan Horse.
Context + Soft Ask. Nothing else.
Claude is going to write this for you.
But Claude naturally writes like an Oxford professor.
You must explicitly command a casual, peer-to-peer tone.
The Prompt: The Trojan Horse
Copy and paste this into Claude:
Act as a top-tier B2B growth hacker.
I am going to give you a prospect's LinkedIn profile data.
Write a 300-character connection request using the 'Trojan Horse' framework.
Rules:
→ 1 sentence establishing context (reference a specific detail from their profile).
→ 1 sentence with a soft, non-salesy reason to connect.
→ Tone must be blunt, casual, and peer-to-peer.
→ Do NOT use greetings like 'Hi' or 'Greetings'.
→ Do NOT pitch anything.
→ Do NOT use the word 'alignment' or 'explore'.
Here is the profile data: [Insert Data]
Skill 8: The Dynamic Multi-Step Sequence Generator
They accepted. Now what?
Most people immediately send a 500-word pitch.
This triggers instant regret in the prospect.
Instead, we run a contextual multi-step sequence.
It adapts based on the exact profile data we extracted.
We use the 3-Step Sequence Template.
→ Day 1: Context (Thank you + relevant observation)
→ Day 3: Value Drop (Send a Lead Magnet related to their pain point)
→ Day 7: Soft Opt-Out (Permission to close the loop)
The Prompt: The Sequence Generator
Feed this to Claude after they accept:
Act as a cutthroat SDR.
Generate a 3-step DM sequence for this prospect based on their profile.
Follow this exact pacing:
Day 1: Casual thanks for connecting. Mention a specific project they are working on.
Day 3: Offer a free resource (Lead Magnet) that solves a problem relevant to their industry. Ask if they want the link.
Day 7: The break-up message. Assume they are busy. Give them an easy out.
Rules:
→ Max 3 sentences per message.
→ Use 8th-grade reading level.
→ No corporate jargon. No pleasantries.
→ Format with heavy white space.
The AI Builder Example
Let's look at this in the wild.
You are targeting a solo founder building AI agents.
You scrape their profile.
You feed it into Claude with Skill 7.
Claude spots a specific GitHub repo linked in their 'About' section.
Or it notices they ranted about Llama 3 in a recent post.
The output looks like this:
"Saw your recent post breaking down the Llama 3 context window limits. Brutal stuff. Building a similar agent stack right now. Let's connect."
It is flawless. It proves you did the work.
Except Claude did the work for you.
Troubleshooting & Pitfalls
Do not let the AI hallucinate familiarity.
If it sounds weird, do not send it.
→ Never pitch in the connection request.
→ Never use "I noticed we are both in..."
→ Never let Claude use words like "Greetings" or "I hope this finds you well."
Keep it raw. Keep it lowercase if that fits your brand.
Treat it like a text message to a coworker.
The Quick-Start Hack
Skip the complex builds if you are short on time.
Do this right now.
Copy your prospect's entire 'About' and 'Experience' sections.
Paste it into Claude.
Fire this exact command:
"Give me 3 casual, non-salesy conversation starters based on this profile. Maximum 2 sentences each. Do not pitch anything."
Pick the best one. Send it.
Watch your reply rate spike.
Bonus: The AI Objection Handling Framework
When they reply, they will have objections.
"Not right now." "We do this in-house." "Too expensive."
Do not freeze. Automate the rebuttal.
Keep this prompt ready in your Claude workspace.
The Prompt: The Rebuttal Engine
My prospect just replied with this objection: [Insert Objection].
I need 3 different ways to reply.
→ Option 1: The 'Agree and Pivot' (Validate their concern, ask a low-friction question).
→ Option 2: The 'Case Study Drop' (Softly mention a similar client who had the same concern).
→ Option 3: The 'Walk Away' (Politely pull the offer back to trigger FOMO).
Keep all replies under 40 words. Conversational tone.
You no longer have to stress over what to say.
You just delegate the thinking to the machine.
Part 5: Closing the Loop: Lead Triage & Analytics Breakdown (Skills 9-10)
You built the scraping engine. You triggered the outbound.
Now you have a new problem.
Your inbox is flooded.
If you spend two hours a day manually reading and sorting LinkedIn DMs, you failed.
Manual triage kills momentum. It destroys your ROI.
We delegate this to Claude.
Skill 9: The Inbound Message Analyzer & Intent Scorer
You need a system to categorize unstructured replies at scale.
Claude acts as your ruthless triage nurse.
It reads the inbox, scores the intent, and triggers the next step in your automation sequence.
→ Hot: "Let's talk." Triggers your calendar link.
→ Warm: "Send me more info." Triggers lead magnet delivery. (I use LeadPanther to automate this exact step — it captures the lead and fires the DM instantly.)
→ Cold: "Not right now." Triggers a 3-month follow-up task in your CRM.
→ DNC: "Stop messaging me." Triggers an immediate blacklist tag.
If you process 100+ DMs a week as a solo ops founder, this skill alone saves your sanity.
You stop wasting mental energy on tire-kickers.
You only spend time talking to prospects who are ready to buy.
The Lead Intent Scoring Matrix (Prompt Template)
Drop this into Claude or your automation webhook:
Role: You are an expert B2B Sales SDR.
Task: Analyze the following inbound LinkedIn message and categorize its intent.
Rules:
- You must output ONLY ONE of these four tags: [HOT], [WARM], [COLD], [DNC].
- [HOT] = Explicit request for a meeting, pricing, or a call.
- [WARM] = Request for more information, a resource, or asking a clarifying question.
- [COLD] = "Not right now," "Maybe later," or "Check back in Q3."
- [DNC] = "Remove me," "Unsubscribe," or any hostile response.
Message: {Insert Prospect Reply}
Troubleshooting the Matrix
Do not give Claude vague instructions.
If you don't define the tags precisely, the AI hallucinates.
It will see "Remove me from your list immediately" and tag it as HOT because the prospect replied quickly.
Define the exact parameters. Audit the first 50 outputs manually to calibrate the prompt.
Skill 10: The Campaign Data Analyst
Scoring leads is half the battle.
The other half is auditing your campaign performance.
You need to know which variables actually move the needle for your business.
Stop staring at messy LinkedIn export CSVs trying to find a pattern.
Feed the raw data into Claude and let it find the winning sequences.
The CSV Analysis Prompt
Export your campaign data and feed it to Claude:
You are a performance marketing analyst.
I am attaching a CSV of my LinkedIn outreach campaign data.
Analyze it and tell me:
1. Which connection request variation had the highest acceptance rate?
2. Which DM sequence step had the highest reply rate?
3. What day of the week performs best for outreach?
4. Which ICP segment converts at the highest rate?
Output your findings as a ranked table. Then give me 3 specific recommendations to improve next week's campaign.
You just replaced a $5,000/month marketing analyst.
The machine sees patterns you miss.
What Comes Next
You have the 10 skills. You have the prompts. You can absolutely build this yourself.
Here is what it actually takes: 2-3 weeks of focused work to set up the full pipeline — Projects configured, Boolean strings tested, DM sequences calibrated, and the triage system wired into your CRM.
The skills work. The execution is where most solo founders stall. You get stuck choosing between building the machine and actually running your business. The prompts sit in a folder. The leads don't get scraped. The DMs don't get sent.
That is the gap between knowing the system and running the system.
Four Ways to Keep Building
1. Start Capturing Leads Today
The skills above are useless without a capture net. When someone comments on your LinkedIn post, LeadPanther auto-captures their profile data and fires a DM — no manual work. Even without the full skill stack, LeadPanther alone will 10x your lead capture from LinkedIn.
2. Join the Community
Agent J is where solo founders building AI automation systems share workflows, troubleshoot prompts, and steal each other's best ideas. The exact skill files from this guide are available inside.
3. Go Deeper with GetDeals
If you want the outbound piece handled — cold email infrastructure, domain warming, mailbox rotation, and AI-written sequences — GetDeals.ai is the system I built for exactly that.
4. Book a Strategy Call
You want someone to wire this entire system for you — LeadPanther, the skill stack, the CRM, the outbound engine — and hand you the keys.
Stop tinkering. Start building.