How the System Works: Claude Creates Attention, Prosp Converts It
Sources:
- Claude Cowork for LinkedIn lead mag (7 LinkedIn skills, setup, DM strategy, lead qualifier with Notion MCP)
- LinkedIn in 2026: The New Strategy for B2B Founders lead mag (three content moats, algorithm system, thought leader ads)
- LinkedIn Hook and Post Template Library lead mag (50+ hook and post formulas)
- PCC: Lara Acosta's LinkedIn System lead mag (positioning, educational and storytelling content, conversion)
- 10K to 6M Impressions lead mag (profile optimisation, content calendar, daily engagement system)
- Claude Code for B2B Cold Email lead mag (personalisation waterfall, list building, warm DM sequencing)
This is the full-stack system for GTM engineers and B2B founders who want LinkedIn to generate pipeline. Not followers for vanity. Pipeline.
Two tools do the work:
- Claude (chat, Cowork, or Claude Code) handles everything that requires writing, research, and analysis: positioning, content creation, post drafting, content analysis, lead magnet ideation, DM copy, lead qualification
- Prosp.ai handles everything that requires outreach at scale: LinkedIn outreach sequencing, DM sending, connection requests, warm follow-up, signal-based prospecting
Five stages:
- Foundation: positioning, profile, CLAUDE.md, and skills installation
- Content: writing posts with Claude that build an audience of your ICP
- Capture: lead magnets and comment CTAs that convert impressions to inbound leads
- Qualify: Claude scores who is worth pursuing, writes to your Notion pipeline
- Convert: Prosp activates warm leads with signal-based sequences, warm DMs, and booked calls
Why Claude and Prosp together
Claude creates attention. Prosp converts it.
The problem with most LinkedIn lead generation: people either create content with no outreach motion, or they run outreach with no content warming the audience first. Both have a ceiling.
Content without outreach: you build an audience but have no system to identify and activate the warm leads inside it. Impressions are vanity. Pipeline requires action.
Outreach without content: cold messages to strangers who have never seen your name. Low reply rates. High account risk. No trust built before the first message lands.
Claude plus Prosp closes both gaps. Claude builds the content that warms the audience. Prosp identifies the warm leads inside that audience and sequences them with signal-based outreach before they go cold. The combination produces inbound and outbound working in the same flywheel.
Stage 1: Foundation: Positioning, Profile, CLAUDE.md, and Skills
Sources: Claude Cowork for LinkedIn lead mag + PCC lead mag
Do all of this before writing a single post. The foundation determines what every post you write will reinforce. Without it, posts feel disconnected and the algorithm does not know who to show you to.
Step 1: Answer the three positioning questions
From Lara Acosta's PCC system. Answer these before writing anything:
- How do you want to be known? One-sentence association you want to own in your ICP's mind.
- Who do you want to be known by? Specific: titles, company types, problems they have.
- What do you want to be known for? One thing. Not a list of services.
Every post you write must reinforce all three simultaneously. This is positioning.
Step 2: Profile optimisation
Do this before posting anything:
Switch primary action from Connect to Follow: Home > Me > Settings and Privacy > Visibility > Followers > Make Follow Primary. Every post shows a follow button inline. Without this, strangers see Connect, which most will not click.
Five sections:
- Profile photo: headshot, approachable, not a logo
- Headline: positioning statement (who you help, what outcome, credibility signal). Not a job title. Keywords throughout.
- About: first person, professional story, point of view established early, keywords throughout
- Featured: pin your lead magnet first, then best posts, then case studies
- Keywords: 5-7 terms that appear in how your ICP searches on LinkedIn. Woven through headline, about, and job titles.
Profiles with complete information get 30% more weekly views (LinkedIn data).
Step 3: Build your CLAUDE.md
CLAUDE.md is the context file Claude reads at the start of every session. Without it, you re-explain your brand, ICP, and voice every single time. With it, every Claude session starts already knowing who you are.
Use this prompt to build it:
"Build my CLAUDE.md. Ask me one question at a time about:
1. My ICP: exact titles, company types, sizes, buying signals
2. My offer in one sentence
3. My proof points with specific numbers
4. My voice: how I write, what I never say
5. My content pillars: the 3-5 topics I talk about with authority
6. My banned phrases and words
7. My LinkedIn goals: followers, leads, calls, revenue
Save the output as CLAUDE.md in my project folder."
Once CLAUDE.md exists, every future Claude session reads it automatically. You never re-explain context again.
CRITICAL: add these rules to your CLAUDE.md regardless of anything else:
## Absolute content rules
- No em dashes anywhere. Ever.
- No hashtags in post bodies
- No banned words: leverage, synergy, bandwidth, genuinely, honestly, straightforward
- No fabricated metrics
- British English
- One idea per line
- No walls of text
- Prosp.ai mentioned once per post maximum if mentioned at all
Step 4: Install the 6 LinkedIn skills in Claude Cowork
Skills are instruction files that load automatically when a relevant task begins. They encode your best practices so you do not brief Claude on them every session.
The 6 skills this system uses:
| Skill | What it does | Trigger |
|---|---|---|
| linkedin-viral-writer | Writes posts using funnel stage, hook formula, brand voice | "Write a LinkedIn post about" |
| lead-magnet-generator | Generates lead magnet ideas that complement a post | "Generate lead magnet ideas for" |
| content-analyser | Analyses posts for hook strength, flow, CTA, improvement | "Analyse this LinkedIn content" |
| warm-dm-strategy | Writes warm DM sequences for engaged prospects | "Write a DM for a first touch" |
| lead-qualifier | Scores a LinkedIn profile against your ICP, writes to Notion | "Qualify this lead" |
| linkedin-to-x | Converts LinkedIn posts to X thread format | "Convert this to X" |
All 6 skills load your CLAUDE.md and who-i-am.md automatically. Every output follows your brand voice without re-briefing.
Install in Cowork: Cowork sidebar > Customize > Skills > Upload each skill .md file. Or install via Claude Code:
npx skills add [skill-name] --global
Stage 2: Content: Writing Posts That Build an ICP Audience
Sources: Claude Cowork for LinkedIn lead mag + PCC lead mag + 10K to 6M lead mag + LinkedIn Hook and Post Template Library
Claude writes every post. Your job is to supply the specific experience, the real numbers, and the genuine insight. Claude handles the structure, the hook selection, and the formatting. You provide what only you can provide.
The two post types and when to use each
From Lara Acosta's PCC system:
Educational: for the logical audience who wants frameworks, steps, and proof you know what you are talking about. Solve one specific problem with your specific experience and proof the solution works. So sharp it feels like reading their mind.
Storytelling: for the emotional audience who needs to see themselves in you before they trust you. Not a fluffy story. Proof of concept through emotion. Before and after. Transformation that shows what is possible.
The content moat check (run before every post):
- Narrative, data, or physical moat present?
- Could AI have written this without your specific experience?
- Is there one specific detail (name, number, date, quote) only you could include?
If question 2 is yes: add one concrete detail before posting.
The post writing workflow
# Step 1: Brain dump your idea (voice note or rough notes)
"I want to write a LinkedIn post about [RAW IDEA].
Here is the specific experience or insight behind it: [DUMP YOUR THOUGHTS]
Here is the proof or result: [SPECIFIC NUMBER OR EXAMPLE]"
# Step 2: Claude writes it (linkedin-viral-writer skill triggers automatically)
# Claude selects the funnel stage, hook formula, and structure
# Returns a draft with: funnel stage, hook formula used, the post
# Step 3: You add the one specific detail only you could write
# Check: is there a name, number, or anecdote that could not have come from AI?
# If no: add it before hitting post
# Step 4: Content analyser skill (optional, run after posting)
"Analyse this LinkedIn content: [PASTE YOUR LIVE POST]"
# Scores hook strength, conversation flow, value delivery, CTA
# Claude saves patterns to memory.md so future posts apply what you learned
Batch creation: block 2-3 hours once a week. Write all posts in one session. Schedule via Buffer or LinkedIn native scheduler. Best times: weekdays 9-5, Tuesday and Wednesday morning.
Posting frequency: 2-3x per week minimum. 2-5x per week optimal. Once a week is still better than zero.
The daily 15-minute engagement system
From the LinkedIn in 2026 lead mag:
20 outbound connection requests per day to ICP only. No message attached. When they accept, they see your next 1-2 posts. 100 per week seeds 100 ICP-relevant people into your audience.
5-10 comments per day on ICP-adjacent content. LinkedIn comments now get more impressions than posts on some accounts. Do not say great point. Add something the original post did not say.
Reply to every comment on your own posts within 1 hour of posting. Buffer data: replying to comments drives 30% more engagement per post. Early engagement velocity tells the algorithm whether to push the post further.
Thought leader ads (when ready to spend): $30/day on best-performing organic posts. Build a company list in Apollo matching your ICP. Upload to LinkedIn Ads for targeting. The PS line URL trick: run posts organically for a week without a link, then edit to add a PS line with the URL before boosting.
Stage 3: Capture: Lead Magnets, Comment CTAs, and Prosp DM Sending
Sources: Claude Cowork for LinkedIn lead mag + PCC lead mag
The comment CTA mechanic is the highest-converting lead gen motion on LinkedIn. Every person who comments has self-identified as interested. The DM arrives to someone who asked for it. Claude writes the post and the lead magnet. Prosp handles the DM sending at scale.
The comment CTA mechanic
How it works: you write a post about a problem your ICP has, offer a resource at the end, and ask them to comment a keyword to receive it. Every comment is an algorithm signal (good for reach). Every commenter self-qualifies as interested. You DM the resource and open a conversation.
The three lead magnet post formulas:
# Formula 1: BIG RESULTS, SIMPLE MECHANISM
[Impressive result] in [short timeframe].
Without [common prerequisite].
Without [common prerequisite].
Without [common prerequisite].
Here's the exact system:
- [Component 1]
- [Component 2]
- [Component 3]
- [Component 4]
- [Component 5]
Comment "[KEYWORD]" and I'll DM it to you.
(Must be following so the DM goes through)
# Formula 2: AFTER X YEARS OF TESTING
After [timeframe] of testing [thing]:
Here's what I know for certain:
- [Contrarian or specific insight]
- [Contrarian or specific insight]
- [Contrarian or specific insight]
I documented the full system.
Comment "[KEYWORD]" and I'll send it.
# Formula 3: THE MISTAKE EVERYONE MAKES
The biggest mistake [ICP] make with [topic]:
[Mistake in 1-2 sentences]
Here's what to do instead:
- [Step]
- [Step]
- [Step]
I built a full [lead magnet type] around this.
Comment "[KEYWORD]" and it's yours.
Lead magnet ideation with Claude
Use the lead-magnet-generator skill after writing any post:
"Generate lead magnet ideas for this content: [PASTE POST]
My ICP: [PASTE FROM CLAUDE.md]
My offer: [WHAT I SELL]
For each idea:
- Title and one-sentence description
- Estimated creation time
- Best delivery method (DM, landing page, email opt-in)
- CTA to embed in the companion post
Rank by: likelihood to attract my ICP and ease of creation."
For interactive tools (highest conversion rate): use Claude Code to build the lead magnet as an interactive calculator or diagnostic tool. One URL that prospects fill in themselves. Prosp can pre-fill their company URL in the link before sending.
# Claude Code: reverse lead magnet build prompt
"Build me a micro SaaS web application:
Purpose: [WHAT THE TOOL DOES FOR MY ICP IN ONE SENTENCE]
Input: [URL OR COMPANY NAME, pre-fillable via ?url= query parameter]
Output: [WHAT THEY SEE: a report, a score, a set of recommendations]
CTA: [BOOK A CALL / JOIN WAITLIST / GET FULL REPORT]
Include email capture before showing full results.
Deploy to Vercel. Use Supabase for email capture.
Before building: ask me clarifying questions."
Prosp for comment-to-DM at scale
Every person who comments your keyword needs a DM. Manually this is unsustainable at volume. Prosp handles the sending.
Workflow:
- Post goes live with comment CTA
- Comments come in
- Prosp detects comments on your posts matching the keyword
- Claude writes the DM template (warm-dm-strategy skill)
- Prosp sends the resource DM automatically
- Prosp queues a follow-up 24 hours later if no reply
- Warm leads flagged for human review
Claude prompt for the resource DM:
"Write the DM to send when someone comments [KEYWORD] on my post about [TOPIC].
Touch 1 (automated, immediate):
- Reference their comment specifically
- Deliver the resource with a direct link or attachment
- One soft question to open the conversation (not a pitch)
- Under 150 characters
Touch 2 (if no reply after 24 hours):
- Short follow-up, different angle
- No pitch
- Under 100 characters
Rules: no em dashes, peer-to-peer tone, sounds like a real person not an automation."
Stage 4: Qualify: Claude Scores Leads and Writes to Notion
Sources: Claude Cowork for LinkedIn lead mag (lead qualifier skill + Notion MCP) + Claude Code for B2B Cold Email lead mag (ICP scoring, signal processing)
Most people who engage with your content are not your ICP. Claude qualifies them so Prosp only pursues the ones worth the effort.
The lead qualifier skill
The skill reads a LinkedIn profile URL, scores the person against your ICP criteria, assigns a heat score, and writes qualified leads directly to your Notion pipeline.
"Qualify this lead: [LINKEDIN URL]
My ICP:
- Titles: [LIST: Director, Founder, Owner, Manager, C-suite]
- Company type: [B2B SaaS / agency / enterprise]
- Company size: [50-500 employees]
- Geography: [UK, US, EU]
- Function: owns or influences outbound, lead gen, or GTM
Disqualifiers:
- Open to work badge (job seekers)
- Company size below 20 employees
- Students, pure IC contributors, non-decision-makers
Return:
- Name, title, company, industry, company size, geography
- ICP match score (1-10) with reasoning per criterion
- Heat score: hot (reach out today) / warm (this week) / cold (monitor)
- Recommended first DM angle based on their profile
- Write to my Notion pipeline: [NOTION DATABASE URL]
Fields: Name, LinkedIn URL, Company, Score, Heat, Status = New Lead, Date Added"
Notion MCP setup: Cowork sidebar > Customize > Connectors > Notion. Authorise with your Notion account. Claude can then write directly to any Notion database you have access to.
Bulk qualification from Prosp comment engagers
After a post drives comments, you have a list of LinkedIn profiles. Run them through Claude Code for bulk qualification before Prosp sequences them:
"Read the CSV of LinkedIn URLs from my post engagers: [FILE]
For each profile, qualify against these ICP criteria:
- Titles: [LIST]
- Company size: [RANGE]
- Function: owns or influences outbound, lead gen, or GTM
- Disqualifiers: job seekers, students, wrong company type
Return a scored CSV with columns:
Name, LinkedIn URL, Company, Title, Size, ICP Score (1-10), Heat (hot/warm/cold), Recommended action
Output hot leads separately as hot-leads.csv
Output warm leads as warm-leads.csv
Skip cold leads entirely."
Hot-leads.csv then goes into Prosp as the sequence audience. Warm-leads.csv goes into a nurture sequence. Cold leads do not get sequenced.
Stage 5: Convert: Prosp Sequences, Personalised DMs, and Booked Calls
Sources: Claude Cowork for LinkedIn lead mag (warm DM strategy) + Claude Code for B2B Cold Email lead mag (personalisation waterfall) + LinkedIn in 2026 lead mag
Prosp activates the warm audience that Claude has built. Every lead entering Prosp sequences has already seen your content, engaged with a post, or commented on a lead magnet. The reply rates are fundamentally different from cold outreach to strangers.
The warm DM sequence: Claude writes, Prosp sends
Claude writes all DM copy using the warm-dm-strategy skill. The three-touch structure:
# Claude prompt for warm DM sequence
"Write a 3-touch LinkedIn DM sequence for warm leads.
Context:
- They engaged with my post about: [TOPIC]
- Their profile suggests they work on: [INFERRED FROM QUALIFIER OUTPUT]
- My offer: [ONE SENTENCE]
- Social proof: [ONE CLIENT RESULT OR CASE STUDY]
Touch 1 (within 24 hours of engagement):
- Reference their specific engagement, not generic
- One sentence on what their profile or engagement suggests about their situation
- Soft open, no pitch, ends with a question about them
- Under 150 characters
Touch 2 (3 days later, if no reply):
- Different angle, relevant insight or observation
- One qualifying question: how long have you been working on that / what has been the hardest part
- Under 200 characters
Touch 3 (7 days after Touch 2, if no reply):
- Honest breakup message, leaves the door open
- Under 100 characters
Rules:
- No em dashes
- No corporate language
- Sound like a peer who noticed something useful
- Move from interest to conversation, not interest to sale
- Never pitch in message one"
Prosp sequence setup
After Claude writes the DM copy and qualifier outputs hot-leads.csv:
- Import hot-leads.csv into Prosp as the sequence audience
- Create a sequence in Prosp with Touch 1, Touch 2, and Touch 3 from Claude's output
- Set delays: Touch 2 after 3 days, Touch 3 after 7 days from Touch 2
- Default settings: stop on reply, LinkedIn native sending limits
- Review before activating
Personalisation waterfall inside Prosp: instead of generic DMs, use the signal Claude identified per lead in the qualifier output. The DM opener references the specific signal (their job posting, their recent content, their tech stack, their growth stage) for each person. Claude Code can generate personalised opener lines per row:
"Read hot-leads.csv.
For each lead, visit their LinkedIn profile and find one personalisation signal:
- A specific post they published recently
- A job posting that indicates a growth area
- A company milestone visible on their page
- Their stated specialisation in their about section
For each lead, write a personalised Touch 1 opener line (under 80 characters)
that references their specific signal and connects to [YOUR OFFER AREA].
Add as a new column: personalised_opener
Save as hot-leads-personalised.csv"
Converting conversations to calls
From Lara Acosta's PCC system: the conversation should move from interest to conversation before it moves to sale.
When they reply positively to Touch 1 or Touch 2, Claude drafts the follow-up:
"They replied to my Touch [1/2] with: [PASTE THEIR REPLY]
My offer: [WHAT I SELL]
My goal: book a 15-minute discovery call
Write the next message in the conversation.
Rules:
- Find the frustration they mentioned casually or the goal they are working towards
- Ask one follow-up question to deepen understanding: 'how long have you been working on that?' or 'what has been the hardest part so far?'
- Do not introduce the offer yet unless they explicitly asked about what I do
- Under 200 characters
- No em dashes"
# Only after 2-3 exchanges and confirmed problem:
"They have confirmed [SPECIFIC PROBLEM].
My offer solves this by [HOW IN ONE SENTENCE].
Write the message that offers a 15-minute call.
Soft, peer-level, specific to what they described.
Under 150 characters."
Quick Reference: The Full System, All Prompts, and Skills
The full system in one place. No em dashes anywhere.
The full system: all 5 stages
# STAGE 1: FOUNDATION (do once, update monthly)
# Positioning
# Answer: how to be known, who by, what for (one thing each)
# Profile
# Follow button: Settings > Visibility > Followers > Make Follow Primary
# Headline: positioning statement, keywords, not a job title
# Featured: lead magnet pinned first
# CLAUDE.md
"Build my CLAUDE.md. Ask me one question at a time:
1. ICP: titles, company types, sizes, buying signals
2. Offer in one sentence
3. Proof points with specific numbers
4. Voice: how I write, what I never say
5. Content pillars: 3-5 topics I own
6. Banned phrases
7. LinkedIn goals
Save to CLAUDE.md"
# Skills install (Claude Cowork)
# linkedin-viral-writer
# lead-magnet-generator
# content-analyser
# warm-dm-strategy
# lead-qualifier
# linkedin-to-x
---
# STAGE 2: CONTENT (weekly)
# Write posts
"Write a LinkedIn post about [TOPIC].
Specific experience behind it: [YOUR EXPERIENCE]
Proof or result: [SPECIFIC NUMBER OR EXAMPLE]"
# linkedin-viral-writer skill triggers, returns draft with funnel stage and hook formula
# Content moat check before posting
# 1. Narrative, data, or physical moat present?
# 2. Could AI have written this without your experience?
# 3. One specific detail only you could include?
# After posting
"Analyse this LinkedIn content: [PASTE POST]"
# Patterns saved to memory.md for future posts
# Daily 15 minutes
# 20 connection requests to ICP only, no message
# 5-10 comments on ICP-adjacent content (add something, not 'great post')
# Reply to all own-post comments within 1 hour
---
# STAGE 3: CAPTURE (2x per month minimum)
# Lead magnet post
"Write a lead magnet post about [TOPIC].
Use Formula 1: BIG RESULTS, SIMPLE MECHANISM
Comment keyword: [KEYWORD]"
# Lead magnet ideation
"Generate lead magnet ideas for this content: [POST]
My ICP: [FROM CLAUDE.md]
Rank by ICP likelihood and ease of creation."
# Prosp: set up comment-keyword detection
# Comments trigger auto-DM (resource delivery + soft open)
# Claude writes all DM copy via warm-dm-strategy skill
---
# STAGE 4: QUALIFY (after every post with comment CTA)
# Single lead
"Qualify this lead: [LINKEDIN URL]
ICP: [FROM CLAUDE.md]
Disqualifiers: [JOB SEEKERS / WRONG SIZE / WRONG INDUSTRY]
Heat score: hot / warm / cold
Write to Notion pipeline: [DATABASE URL]"
# Bulk from post engagers
"Read [engagers.csv].
Qualify each against ICP criteria.
Output hot-leads.csv, warm-leads.csv.
Skip cold leads."
---
# STAGE 5: CONVERT (ongoing, Prosp runs the sequences)
# Claude writes the DM sequence
"Write a 3-touch LinkedIn DM sequence for warm leads
who engaged with my post about [TOPIC].
Touch 1: within 24 hours, reference their engagement, soft open
Touch 2: 3 days later, qualifying question
Touch 3: 7 days later, honest breakup
No em dashes. Peer-level tone. Under 150 characters each."
# Personalised openers (Claude Code)
"Read hot-leads.csv.
Visit each LinkedIn profile.
Write one personalised opener (under 80 characters) per lead
referencing their specific signal.
Save as hot-leads-personalised.csv"
# Import hot-leads-personalised.csv into Prosp
# Sequence: Touch 1 personalised opener > Touch 2 (day 3) > Touch 3 (day 10)
# Stop on reply. Review before activating.
# When they reply: Claude drafts next message
"They replied: [PASTE REPLY]
My offer: [ONE SENTENCE]
Goal: book a 15-minute call
Next message: one qualifying question, no pitch yet
Under 200 characters. No em dashes."
Skills quick reference
# LINKEDIN VIRAL WRITER
Trigger: "Write a LinkedIn post about [TOPIC]"
Output: post with funnel stage and hook formula shown
# LEAD MAGNET GENERATOR
Trigger: "Generate lead magnet ideas for this content: [POST]"
Output: ranked ideas with delivery method and CTA per idea
# CONTENT ANALYSER
Trigger: "Analyse this LinkedIn content: [POST]"
Output: hook score, flow, value delivery, CTA strength, one biggest improvement
# WARM DM STRATEGY
Trigger: "Write a DM for a first touch" or "Write a 3-touch DM sequence"
Output: Touch 1/2/3 with character counts
# LEAD QUALIFIER
Trigger: "Qualify this lead: [LINKEDIN URL]"
Output: ICP score, heat score, recommended DM angle, writes to Notion if qualified
# LINKEDIN TO X
Trigger: "Convert this to X" + paste post
Output: numbered thread, hook tweet standalone, one idea per tweet
What to build first
Week 1: CLAUDE.md, profile with Follow Primary on, featured section with lead magnet pinned. Install skills.
Week 2: First 3 posts. Run content analyser on each after posting. Start daily 15-minute engagement routine.
Week 3: First lead magnet post with comment CTA. Set up Prosp comment-to-DM automation. Run qualifier on first batch of commenters.
Week 4: First Prosp sequence live with hot leads from qualifier. Claude drafts all DM copy. Prosp sends.
From week 4 onwards: the flywheel runs. Content builds the audience. Lead magnets capture interest. Claude qualifies. Prosp converts warm leads to calls.
