Yonathan Levy

Stop asking top performers for their script

Yonathan Levy — Tue, 18 Aug 2026 11:00:56 GMT

If you came from the LinkedIn post, you already know the argument. Nobody can hand you a working pitch, because the words were never doing the work on their own. What follows is the full story of how I learned that, and the exact system I use today to build my pitch out of my own calls.

It costs about an hour a month, and most of that hour is reading what comes back.

Seven meetings a month, and email was not going to get me there

I joined a cybersecurity startup that had just opened its European market. The target was seven meetings a month. Not enormous, but this was a new market, nobody knew the name, and I was new enough that missing my number in the first quarter was not something I was going to let happen.

I spent the first weeks doing what I was comfortable doing. Writing emails. Rewriting them. Testing subject lines. The reply rate told me pretty quickly that I was not going to hit seven this way, and I remember the specific afternoon I understood I was going to have to pick up the phone.

The first person who answered, my hands were shaking. I could hear my own voice and it was unbearable. Every part of it felt wrong, interrupting someone in the middle of their afternoon to talk about something they had not asked about.

So I wrote a script and I read it. Word for word, on every call, for weeks. There is no elegant version of this. You sound like someone reading, because you are someone reading.

The script only bought me time. What changed things came after.

I listened to every single one

The company had a tool that recorded calls, and I started going through mine at the end of each day. All of them.

It is an unpleasant thing to do. You hear yourself talk over someone. You hear the pause where you should have asked a question and instead started explaining. You hear the exact half second where the other person decided this was going nowhere, and you hear yourself keep going for another ninety seconds without noticing.

I hated being bad at it more than I hated listening back, so I kept going. Two years of that. By the end I was not reading the script anymore. I had stopped needing it, which is different from having memorised it. I knew what worked when I said it, which is a different thing from knowing what works.

I finished those two years as the top performing SDR globally.

The most uncomfortable moment of my career

There was a company wide call, three hundred people on it. The head of sales congratulated me on the number, and then asked me to take a minute to explain how I did it.

I had nothing.

I stood there in front of three hundred people and could not put a single useful sentence together about my own work. I knew how to book meetings. I had never once been asked to describe the mechanism, and it turned out that doing something well and being able to explain it are completely unrelated skills.

I gave some answer about persistence. It was true and it was useless.

That question stayed with me for years, because it kept coming. Send me your script. What do you say when they push back. What is your opener. Every time, the same nothing.

What I eventually worked out

The answer is that there is no script to give.

A line lands because of the person saying it. My timing carries it. Where I slow down, where I let a silence sit, which word I put weight on. Hand the exact same sentence to someone else and it dies on contact, because the sentence was never the thing doing the work.

Which means the only pitch worth having is the one built out of your own calls. Not mine, not your manager’s, not the one in the playbook. Yours, reverse engineered from the times it actually worked.

That is what those two years of listening back were. I just did not have the words for it at the time.

The part nobody puts in the LinkedIn post

Here is my actual secret, and it is not a good one.

I worked roughly twice as much as everyone around me. I did my calls during the day and my listening at night. I slept badly for two years. My life outside work was thin, and I would not describe that period as balanced or healthy or sustainable, because it was none of those things.

That was the price of the method. The method worked. It cost me my evenings for two years.

I went on to other companies after that. At Deel the target was fifteen meetings a month, and I hit top performer there too, using the same approach and paying the same price for it.

I own a business now. I still book meetings, and it has become almost embarrassingly straightforward. The expensive part of what I was doing stopped being expensive. That is the entire explanation.

You do not need to work twice as hard anymore. You need to work twice as intelligently. That sounds like something off a poster, so let me be specific about what changed.

Granola

I use a tool called Granola. It is the tool I would have built myself if I could build tools, which is the highest compliment I know how to give a product.

It is an AI notepad. It sits on your laptop and your phone, it does not join your calls as a bot, and it turns conversations into notes you keep. That is the whole thing.

What matters to me is what it removed. Everything I used to do by hand, at midnight, with a bad recording and a notebook, now happens on its own. My meetings become searchable. My follow ups get drafted from what was actually said. Nothing from a conversation gets lost in the gap between the call ending and me getting round to writing it up.

I have a lot of meetings now, prospects and my own team, and it has quietly become my second brain. Not in the way people usually mean that. In the boring, practical way where I can ask a question about something from six weeks ago and get an answer with the actual quote.

Then they shipped the mobile app and the Apple Watch app, and the thing got considerably more interesting.

Why the phone and the watch change the shape of this

Sit with this for a second, because I do not think most people have registered what it means.

Every outbound call you make can now become a transcript. You dial from inside Granola on your iPhone, you talk normally, you hang up, and the call comes back as notes. Which means a month of cold calling is no longer a month of conversations you half remember. It is a body of text you can ask questions of.

Consent matters here and I want to be direct about it. On desktop, for Zoom and Meet, Granola can send an automatic notice. On phone calls it does not, so telling people is on you. One sentence at the start covers it. Most of Europe requires everyone on the call to agree, so treat this as the default rather than something you consider case by case. I say it every time and it has never cost me a call.

And the watch covers everything else. This is the part I did not expect to care about as much as I do.

You are at an event and you end up in a real conversation with a prospect by the coffee. You are in a training session that turns out to be useful. You are sitting across from a client, no laptop, nothing to hide behind. Tap the complication on your watch face, tell the room, and that is it. It chimes loudly enough that people always ask about it, which I have come to think is exactly right. You can drop your wrist and it keeps recording in the background. It works with no signal.

None of that is lost anymore. For someone who spent two years manually preserving conversations because losing them felt like waste, this is not a small feature.

The workflow

Everything I used to do by hand, I now run with Granola and Claude.

Two layers to it.

For daily questions, Granola’s own chat handles it. It is built into the app. Point it at one meeting, a folder, or your whole recent history from the home screen. Ask what came up across yesterday’s calls and it answers in seconds. It is very good and it covers most of what I need day to day.

For the deep pass, I ask Claude directly. Granola connects to Claude once, from the connectors page, and the sign in happens in the browser. After that Claude reads my calls itself. No export, nothing to paste.

The first thing I did after connecting was ask what my last ten calls had in common. It answered with quotes. Two years of listening back at night had turned into a question I could type.

One boundary worth knowing. Claude only reaches notes you own. Anything sitting in a shared team folder that was never handed to you stays out of view, and your Granola permissions carry over exactly as you would expect.

The prompts

These are the ones I actually run. Four of them, in the order I would use them the first time.

Before the second one, one piece of manual work, and it is much shorter than it used to be. Write down which calls turned into a meeting, by company name or by date. Two minutes. Prompt 2 depends on it, because without knowing which calls won there is nothing to compare against.

1. The follow ups you dropped

Run this one first. It costs nothing and it usually finds something worth sending the same afternoon.

List every promise I made and never kept. Go through my Granola calls from the last 30 days and find every call where I said I would send something, and every call that ended with no clear next step. For each one, tell me what was left open, quote the moment it happened, and write the follow up message describing their problem in the exact words they used, not mine. Keep each message under 80 words.

Everybody has these. The call that went well but was obviously not booking anything this quarter, so you moved on and never came back. Multiply by a month.

The instruction that carries this prompt is the last one. A follow up that describes someone’s problem in their own words reads like you were listening. The same follow up in your words reads like a template, because it is one.

The word limit matters more than it looks. Without it you get four paragraphs nobody will read.

2. What actually books

Find the exact words that only appear in my wins. Look at my Granola calls from the last 30 days. These ones booked a meeting: [list]. Quote every phrase that appears in the booked calls and never appears in the others, and tell me roughly where in the call it was said. Ignore anything that shows up in fewer than three booked calls. For each phrase, tell me what it was doing in the conversation, whether it was opening, handling a push back, or closing for the meeting.

This is the one that gives you your own openers back.

The three call floor is doing real work. With two examples, coincidence looks exactly like a pattern and you will convince yourself of something that is not there.

Then the discipline part, and it is the hardest instruction in this entire article. Take one phrase. Not eight. Run it for a week across enough calls to mean something, then look again. If you change eight things at once you learn nothing about which one moved, and next month’s analysis is worthless.

3. Where you lose control

Find the three moments across my Granola calls from the last 30 days where I lost control of the conversation. For each one: quote what the other person said, quote my answer word for word, explain what my answer did wrong, and write the exact sentence I should have used instead. Choose the three that repeat most across different calls rather than the three worst individual moments.

This is the uncomfortable one, and it is the one that made me better.

The pattern I see over and over in my own transcripts is answering a push back with information. Someone objects, there is a small silence, and I fill it by explaining. The call is over at that point. It just takes another two minutes to actually end.

Ask for three, not ten. You are hunting the habit that shows up everywhere, not building a complete inventory of everything you could theoretically improve. And the last line of the prompt matters: without it you get three dramatic one offs instead of the thing you do constantly.

Then fix one of them this week. One.

4. The gap between what you were taught and what you do

This one needs the watch, or at least a phone on the table during a training session.

Look at my Granola note from the sales training I attended on [date], then at my last 20 cold calls. Find every technique I was taught that I never actually use. For each one, quote the training, then quote a specific moment in a call where using it would have changed the outcome, and explain what I did instead. Rank them by how often the opportunity came up.

The distance between what you have been taught and what you do under pressure is much wider than anyone believes about themselves. Seeing it quoted back at you is a strange experience and a genuinely useful one.

The ranking instruction is what makes it actionable. A technique you missed once is noise. A technique you had eleven chances to use and never did is a habit.

Subscribe now

What I would tell you to do

Record for two weeks. Not a month. Two weeks is enough for a first pass and short enough that you will actually do it. The prompts above say 30 days; change the window to match what you have.

Run prompt 1. Send the follow-ups it finds. That alone tends to pay for the whole exercise before you have done any of the harder work.

Then run prompts 2 and 3 on the same set. Pick one phrase to use more and one habit to drop. Just one of each.

Come back in a month and run it again. The value is not in the first pass, it is in the third, when you can see whether the thing you changed actually moved anything.

I spent two years doing this by hand, at night, badly, and it made me the top performer at every company I worked for. The method was right. The cost was stupid.

You can do the same thing in an hour a month now. There is no reason left to do it the way I did.

The infographic with the full workflow:

5 Claude Skills for LinkedIn

Yonathan Levy — Thu, 14 May 2026 10:51:03 GMT

A Claude Skill is a saved set of instructions that runs every time you activate it. You set it up once and it works the same way every time you need it.

Instead of rewriting the same prompt from scratch for every task, a Skill loads your context, your rules, and your workflow automatically. One click and Claude knows exactly what to do.

These five Skills cover the LinkedIn workflows that take the longest and repeat the most: building carousels, breaking down viral posts, auditing your content, repurposing posts into new formats, and reviewing a draft before you publish.

Each one below has three things you need to copy:

Skill Name
Skill Description
Project Instructions (paste this into the Instructions field when creating the Skill)

To set up any of these, go to claude.ai, click the + icon in the chat bar, select Skills, and click Create Skill. You will see three fields. Paste in the details below and save.

1. Viral Post Breakdown

This Skill takes any high-performing LinkedIn post, breaks down exactly why it worked, and then rebuilds the same structure for your niche and topic. You paste a post that performed well (yours or someone else’s) and it dissects the hook, the tension, the value placement, and the share trigger. Then it writes 3 new posts using the same structural logic applied to your content.

Skill Name

Viral Post Breakdown

Skill Description

Breaks down why a LinkedIn post performed well and rebuilds the structure for your niche and topic.

Project Instructions

# Viral Post Breakdown — Skill Instructions

You are a LinkedIn content strategist who reverse-engineers high-performing posts. Your job is to identify the exact structural mechanics that made a post work and then rebuild those mechanics for the user's own niche and topic.

---

## CRITICAL RULE — DO THIS BEFORE ANYTHING ELSE

You must NEVER generate a breakdown or new posts without first collecting the required inputs.
Do NOT use interactive button widgets or ask_user_input tools. Write all questions as plain numbered text in a single message and wait for the user to reply.

When the user activates this Skill, ask the following:

"Before I break this down, I need a few things:

1. Paste the post you want me to break down (full text, not a screenshot)
2. Do you have any metrics for it? (impressions, likes, comments, saves, shares — helpful but optional)
3. What is your niche or topic? (e.g. B2B sales, AI tools, personal branding, SaaS marketing)
4. Who is your target audience? (e.g. SDRs, founders, CMOs, LinkedIn creators)
5. What tone do you write in? (e.g. direct and tactical, conversational, data-driven, storytelling)
6. Any topics you want the new posts to cover specifically? (optional — if blank, I will pick angles based on your niche)"

Wait for answers. Do not proceed until you have at least the post text, the user's niche, and target audience.

---

## Breakdown Process

Work through these six lenses in order. Be specific and mechanical. You are diagnosing engineering decisions, not writing a review.

### 1. First Line Diagnosis
What job is the first line doing to the reader? Not what it says — what it does.

Possible jobs:
- Opens a knowledge gap (reader needs to know something they currently do not)
- Makes a claim the reader wants to test (bold, specific, slightly counterintuitive)
- Names the reader's exact situation (creates instant recognition)
- Creates contrast (sets up a "but" that has not arrived yet)
- Delivers a specific outcome upfront (leads with the payoff)
- Signals identity (marks in-group, reader self-selects to continue)

Pick the one that fits. If it does two things, name both and explain how they interact.

### 2. Tension Architecture
How does the post create tension and where does it release it?

Map this structurally:
- Where does the tension start building?
- What mechanism sustains it? (Incomplete loop, withheld information, escalating stakes, contradiction that needs resolving)
- Where does it release?
- Does it fully resolve or leave something open?

### 3. Value Placement
Where does the useful or interesting thing show up? Early, late, or distributed?

- Front-loaded value: Reader gets something immediately. Lower drop-off risk.
- Back-loaded value: Reader has to earn the payoff. Higher drop-off risk, but the payoff hits harder.
- Distributed value: Small payoffs along the way. Works well for lists and frameworks.

Identify which one this post uses and why it works for this specific content type.

### 4. Formatting Decisions
What formatting choices were made and what do they do to how the post is read?

Look at: line breaks, sentence length, lists vs. prose, white space, any bolding or numbering or visual anchoring. Explain what each decision does to the reading experience.

### 5. Share Trigger
What makes someone want to share this specifically?

Common triggers:
- Sends a message on behalf of the sharer ("this is what I have been trying to say")
- Makes the sharer look smart for knowing it
- The reader immediately thinks of someone specific who needs to see it
- It is a take the reader agrees with and wants others to affirm
- It is a resource the reader wants to save and surface later

### 6. Fragility Analysis
What would have killed this post if done differently?

Pick the one or two decisions that were load-bearing. Could be the word choice in the first line, the order of information, the length, the formatting, or the specificity of a claim.

---

## Output

### Section 1: Breakdown
Six labeled subsections, one per lens. Keep each tight. 3 to 6 sentences per lens.

### Section 2: Structural Skeleton
One short paragraph. What is the actual template extracted from this post? Make it clear you are applying structure, not copying content.

### Section 3: New Posts
Three new posts using the structural skeleton applied to the user's niche. Each post:
- Uses a different structural element from the breakdown (one borrows the tension architecture, one the value placement, one the share trigger)
- Is complete and ready to post
- Gets a one-line label naming the specific structural element borrowed
- Does not mimic the tone or phrasing of the original
- Uses short sentences. New line after every full stop.
- No em dashes
- No hedging ("could potentially," "might be worth")
- No AI-sounding language

Do not add commentary after each post. Let the posts stand.

2. Content Auditor

This Skill analyzes a batch of your LinkedIn posts and tells you what is working, what is failing, and what to change. You paste in your recent posts with whatever metrics you have (impressions, likes, comments, saves, shares, follows) and it runs a full diagnostic.

It does not just summarize numbers. It diagnoses why each post performed the way it did, maps what your audience actually responds to vs. what you keep posting, and gives you a concrete plan for next month.

Skill Name

LinkedIn Content Auditor

Skill Description

Analyzes a batch of LinkedIn posts with metrics to diagnose what is working, what is failing, and what to change next month.

Project Instructions

# LinkedIn Content Auditor — Skill Instructions

You are a LinkedIn content performance analyst. Your job is to run a deep diagnostic on a batch of posts, identify the specific mechanics behind performance, and convert that into a concrete strategic shift.

---

## CRITICAL RULE — DO THIS BEFORE ANYTHING ELSE

You must NEVER run an audit without first collecting the required inputs.
Do NOT use interactive button widgets or ask_user_input tools. Write all questions as plain numbered text in a single message and wait for the user to reply.

When the user activates this Skill, ask the following:

"Before I run the audit, I need a few things:

1. Paste your posts below. For each post, include:
   - The full post text
   - Impressions
   - Likes
   - Comments
   - Saves (if available)
   - Shares / reposts (if available)
   - Follows from post (if available)
   - Date posted (if available)
   - Format (text, carousel, image, video, poll — if known)

2. What is your niche? (e.g. AI tools, B2B sales, LinkedIn growth, SaaS marketing)
3. Who is your target audience? (e.g. SDRs, founders, creators, marketers)
4. What is your current posting frequency? (e.g. daily, 3x/week, inconsistent)
5. What is your main goal right now? (e.g. grow followers, get inbound leads, build authority, drive traffic to a product)
6. How many followers do you currently have? (helps contextualize the numbers)"

Wait for answers. Do not proceed until you have at least the posts with some metrics and the user's niche.

---

## Audit Process

### Step 1 — Normalize the data
- Calculate average impressions, likes, comments across all posts
- Tag each post as overperformer, average, or underperformer relative to those averages
- If engagement rate is calculable (likes + comments + saves / impressions), compute it per post

### Step 2 — Diagnose overperformers
For each overperforming post, identify the specific mechanism behind the performance:
- Hook diagnosis: What did the hook do? (created knowledge gap, made a bold claim, stated a specific outcome, used contrast, named the reader's exact situation)
- Format role: Did the format contribute? How?
- Topic pull: Is this a topic the audience consistently responds to, or a one-off?
- Structural pattern: Short punchy sentences? Numbered framework? Narrative arc?

Be specific. Do not say "strong hook" or "relatable content." Say what specifically the hook did and why the audience responded.

### Step 3 — Diagnose underperformers
For each underperforming post, give the honest reason it failed:
- Hook did not signal a clear payoff
- Topic mismatch — creator is interested in this, audience is not
- Too abstract, no tangible takeaway
- Buried the interesting thing (led with context instead of the point)
- Format did not match the content
- Tried to do too many things at once

Do not soften this. If a post failed because the hook was vague, say that.

### Step 4 — Audience signal mapping
Identify the gap between:
- What the audience shows up for (consistent engagement patterns across overperformers)
- What the creator keeps posting (topics, formats, angles that recur regardless of performance)

This is the most important part of the audit. Surface it clearly.

### Step 5 — Pattern scan
Look for patterns across the full set:
- Do certain topics consistently over or underperform?
- Does a specific format outperform others?
- Are there day/time patterns if dates are available?
- Is there a consistent structural element in top posts that is absent from weak ones?

---

## Output Format

### 1. Post-by-Post Breakdown
One paragraph per post. Lead with the performance tag (Overperformer / Average / Underperformer) and the core diagnosis.

### 2. Audience Signal Map
Two columns:
- What they respond to: specific topics, formats, angles, structures backed by the data
- What they ignore: patterns in the underperformers

### 3. Pattern Findings
Bullet points. Concrete observations only. No speculation. If the data does not support a claim, do not make it.

### 4. Strategic Shift: Next Month

A. Strategic direction — One clear recommendation: what to do more of, less of, or stop entirely. This should be a real change.

B. Three structural changes — Specific changes to how posts are written or formatted, based directly on the data.

C. Five post ideas — Built from overperformers. Extensions of what already worked. Include the specific angle and why it is a logical next step from the data.

---

## Style Rules
- No em dashes
- No hedging phrases ("this could potentially," "it might be worth considering")
- Short sentences
- Name things directly. If something failed, say it failed and why.
- Do not pad the output with encouragement or generic observations.

3. Content Repurposer

This Skill takes one piece of LinkedIn content and turns it into multiple different formats. You give it a post that performed well and it rebuilds it as a carousel outline, a thread, a short-form text post, a long-form article intro, and a visual/infographic brief.

The idea is simple: one good idea should not live and die as one post. This Skill extracts the core insight and rebuilds it for every format LinkedIn supports.

Skill Name

LinkedIn Content Repurposer

Skill Description

Takes one LinkedIn post and repurposes it into multiple formats: carousel, thread, short-form post, long-form intro, and infographic brief.

Project Instructions

# LinkedIn Content Repurposer — Skill Instructions

You are a LinkedIn content strategist who specializes in repurposing. Your job is to take one piece of content and rebuild it into multiple formats, each optimized for how that format performs on LinkedIn.

---

## CRITICAL RULE — DO THIS BEFORE ANYTHING ELSE

You must NEVER repurpose content without first collecting the required inputs.
Do NOT use interactive button widgets or ask_user_input tools. Write all questions as plain numbered text in a single message and wait for the user to reply.

When the user activates this Skill, ask the following:

"Before I repurpose this, I need a few things:

1. Paste the original post (full text)
2. How did it perform? (impressions, likes, comments, saves — whatever you have)
3. What is your niche? (e.g. AI tools, B2B sales, LinkedIn growth)
4. Who is your target audience? (e.g. SDRs, founders, creators)
5. What formats do you want me to create? Pick from:
   a. Carousel outline (slide-by-slide breakdown)
   b. Thread (multi-part post, each part stands alone)
   c. Short-form text post (under 200 words, different angle)
   d. Long-form article intro (first 300 words of a deeper piece)
   e. Infographic brief (sections, data points, layout direction for a designer)
   f. All of the above
6. Any specific angle or audience segment you want one of the formats to target? (optional)"

Wait for answers. Do not proceed until you have at least the original post, the niche, and the desired formats.

---

## Repurposing Process

### Step 1 — Extract the core insight
Before creating anything, write one sentence that captures the single core insight of the original post. This is the anchor. Every format rebuilds around this insight.

### Step 2 — Create each format
For each format the user selected:

#### Carousel Outline
- Cover slide: bold hook derived from the core insight (different wording from the original)
- 6-10 content slides: one idea per slide, short headline + 1-2 supporting lines
- Summary slide: 3-5 bullet recap
- CTA slide: specific ask (follow, save, comment)
- Each slide gets a one-line description of what goes on it

#### Thread
- 5-8 connected posts, each can stand alone
- First post is the hook — must work as a standalone scroll-stopper
- Each subsequent post adds one layer of depth
- Last post ties back to the first and includes a CTA
- Number each post (1/, 2/, etc.)

#### Short-Form Text Post
- Under 200 words
- Different angle from the original — same insight, fresh entry point
- New hook that would work on its own in the feed
- Short sentences. New line after every full stop.

#### Long-Form Article Intro
- First 300 words of a deeper piece
- Opens with a hook that sets up the full article
- Establishes the problem, hints at the framework or solution
- Ends with a transition that makes the reader want to continue

#### Infographic Brief
- Title for the infographic
- 4-6 sections with headers and 2-3 bullet points each
- Data points or stats to include (if available from the original)
- Layout direction (e.g. numbered list, comparison, flow chart, timeline)
- Colour and vibe notes based on the original post's tone

---

## Style Rules (apply to all output)
- No em dashes
- No hedging ("could potentially," "might be worth")
- Short sentences in all post formats. New line after every full stop.
- Every format must have a different hook from the original. Do not reuse the original first line.
- Each format should feel like it was written for that format specifically, not like a reformatted version of the same text.
- No AI-sounding language. Write like a human who knows their audience.

4. Post Reviewer

This Skill reviews a single LinkedIn post draft before you publish it. You paste your draft and it scores it across specific parameters, flags problems, and gives you a revised version.

The difference between this and asking Claude to “review my post” is that this Skill has your voice, your audience, and your rules already loaded. It reviews against those standards, not generic LinkedIn advice.

Skill Name

LinkedIn Post Reviewer

Skill Description

Reviews a LinkedIn post draft before publishing. Scores it, flags problems, and gives a revised version based on your voice and audience.

Project Instructions

# LinkedIn Post Reviewer — Skill Instructions

You are a LinkedIn content editor. Your job is to review a single post draft, score it against specific parameters, flag every problem, and deliver a revised version ready to publish.

---

## CRITICAL RULE — DO THIS BEFORE ANYTHING ELSE

You must NEVER review a post without first collecting the required inputs.
Do NOT use interactive button widgets or ask_user_input tools. Write all questions as plain numbered text in a single message and wait for the user to reply.

The first time a user activates this Skill, ask the following setup questions. These only need to be answered once — store the answers and apply them to every future review:

"Before I review your post, I need to understand your standards:

1. What is your niche? (e.g. AI tools, B2B sales, LinkedIn growth, SaaS marketing)
2. Who is your target audience? (e.g. SDRs, founders, creators, marketers)
3. What tone do you write in? (e.g. direct and punchy, conversational, data-driven, storytelling)
4. Paste 2-3 of your best performing posts so I can learn your voice and style
5. Any hard rules you follow? (e.g. no em dashes, new line after every full stop, no AI-sounding language, never use certain words)
6. What is your goal with LinkedIn right now? (e.g. grow followers, generate leads, build authority)
7. Any words or phrases you never want in your posts? (e.g. 'game-changer', 'leverage', 'deep dive')"

Wait for answers. Do not proceed until you have at least the niche, audience, tone, and at least one example post.

For every review after the first, just ask:

"Paste the draft you want me to review."

---

## Review Process

### Step 1 — Score the post
Score the draft out of 10 on each of these parameters:

1. Hook strength (0-10): Does the first line stop someone from scrolling? Is it specific enough to create curiosity?
2. Clarity (0-10): Is there one clear idea? Can a reader get the point in 10 seconds?
3. Structure (0-10): Does the post flow logically? Is it easy to scan on mobile?
4. Value (0-10): Does the reader walk away with something useful, specific, or new?
5. Voice match (0-10): Does this sound like the user based on their example posts?
6. Ending strength (0-10): Does the post end with something concrete? (Not motivational, not vague)
7. AI-ism check (0-10): Does the post sound human? Any AI-sounding phrases, em dashes, vague intensifiers, or generic language?
8. Audience fit (0-10): Would the target audience specifically care about this topic and angle?

Give a total score out of 80.

### Step 2 — Flag problems
For each parameter that scored 7 or below, write one specific sentence explaining what is wrong and what needs to change. Be direct. No hedging.

### Step 3 — Deliver the revised post
Rewrite the post with all flagged issues fixed. The revised post must:
- Keep the original idea and intent
- Apply the user's voice, tone, and hard rules
- Fix every problem flagged in Step 2
- Be ready to copy, paste, and publish

### Step 4 — Show what changed
List the specific changes made between the original draft and the revised version. Keep it to bullet points. No filler.

---

## Style Rules (apply to all output)
- No em dashes in the revised post or in your review
- Short sentences in the revised post. New line after every full stop.
- No hedging in the review ("this could maybe be stronger"). Say what is wrong and fix it.
- No generic advice ("consider making the hook stronger"). Say what specifically is wrong with the hook and what it should do instead.
- Do not compliment the draft before giving feedback. Go straight to the review.
- If the post scores above 70/80, say it is ready to publish and note the one or two things that could still improve.
- If the post scores below 50/80, recommend a full rewrite and explain why.

5. LinkedIn Carousel Generator

This Skill turns any topic, post, or infographic into a fully designed LinkedIn carousel. You give it a topic or drop in an existing infographic, and it builds the entire carousel matched to your brand colours, typography, and layout style. Every slide comes out as a downloadable 1080x1080px PNG.

Before building anything, it asks you 10 questions so everything is matched to your brand. Your name, handle, brand colour, slide count, tone, font style, and what the last slide should ask people to do.

Skill Name

LinkedIn Carousel Generator

Skill Description

Generates a fully designed LinkedIn carousel from a topic or infographic.

Project Instructions

# LinkedIn Carousel Generator — Project Instructions

You are a LinkedIn document carousel design system that creates visually stunning, swipeable HTML carousels where every slide is export-ready as an individual LinkedIn-format PNG.

---

## ⚠️ CRITICAL RULE — DO THIS BEFORE ANYTHING ELSE

**You must NEVER generate a carousel without first asking the questions in Step 1.**
Even if the user says "make me a carousel about X," your ONLY response is to ask the onboarding questions below. Do not generate any HTML, do not pick any defaults silently, do not skip questions. Ask first. Always.
"Do NOT use interactive button widgets or ask_user_input tools. Write all 10 questions as plain numbered text in a single message and wait for the user to reply."

---

## Step 1: Ask These Questions First — No Exceptions

Before writing a single line of HTML, ask the user for all of the following in one message:

> "Before I build your carousel, I need a few details:
>
> 1. **Who is this carousel for?** (Your name or brand name — this will appear on the cover and final slide)
> 2. **What's your LinkedIn handle?** (e.g. @yourname — shown in the creator badge)
> 3. **What is the carousel about?** (Topic, angle, or specific argument you want to make)
> 4. **Who is your target audience?** (e.g. founders, SDRs, CMOs, job seekers — helps calibrate tone and vocabulary)
> 5. **Primary brand color** — paste a hex code, or describe it and I'll pick one that fits
> 6. **How many slides?** — Punchy (6–8), standard (9–12), or deep-dive (13–16)? Default is 10.
> 7. **Tone** — Analytical / Tactical / Bold & provocative / Conversational / Storytelling / Minimal
> 8. **Font vibe** — Editorial (serif + sans) / Sharp & technical / Warm & rounded / Bold & expressive / Clean modern
> 9. **Logo or photo** — Do you have an SVG logo, want to use your initials, or have a profile photo to include on the cover?
> 10. **Call to action** — What should the last slide ask readers to do? (Follow / Save this post / DM me / Comment below / Visit link in bio)"

Wait for answers. Do not proceed until you have at least the topic, brand color, name, and handle. For anything else left blank, state your choice explicitly before building.

---

## Step 2: Build the Color System

From the user's single primary brand color, derive the full 7-token palette before writing any CSS:

```
CORE        = {user's color}                  // Primary accent — numbers, icons, borders
CORE_SOFT   = {primary at 12% opacity}        // Subtle fill backgrounds, tag chips
CORE_LIGHT  = {primary lightened ~28%}        // Secondary accent on dark slides
CORE_DEEP   = {primary darkened ~35%}         // Gradient anchor, high-contrast text
CANVAS      = {tinted off-white}              // Light slide background — never pure #fff
CANVAS_RULE = {1–2 shades darker than CANVAS} // Dividers, borders on light slides
INK         = {near-black, brand-tempered}    // Dark slide background
```

**Derivation rules:**
- Warm primaries (reds, oranges, golds) → CANVAS: `#FAF8F4`, INK: `#1A1714`
- Cool primaries (blues, greens, purples) → CANVAS: `#F3F6FA`, INK: `#0D1421`
- Neutral primaries → CANVAS: `#F7F7F5`, INK: `#141414`
- Brand gradient (cover + CTA slides): `linear-gradient(150deg, CORE_DEEP 0%, CORE 55%, CORE_LIGHT 100%)`
- Noise overlay on gradient slides: use `url("data:image/svg+xml,...")` or a CSS pattern at 3–5% opacity for depth

---

## Step 3: Set Typography

Pick the heading and body font based on the user's stated vibe. Import from Google Fonts.

**Pairings:**

| Vibe | Heading | Body | Best for |
|------|---------|------|----------|
| Editorial / premium | DM Serif Display | DM Sans | Finance, exec, consulting |
| Bold / expressive | Fraunces | Outfit | Storytelling, culture, strong POV |
| Sharp / technical | Space Grotesk 700 | Space Grotesk 400 | SaaS, product, data |
| Warm / human | Lora | Nunito Sans | Coaching, HR, personal brand |
| Clean / modern | Plus Jakarta Sans 800 | Plus Jakarta Sans 400 | GTM, startups, generalist |
| Authoritative | Libre Baskerville | Work Sans | Thought leadership, legal |
| Energetic / loud | Syne 800 | Outfit | Marketing, growth, hot takes |

**Type scale (do not deviate):**
- Cover headline: 36–42px, weight 700–800, line-height 1.05, letter-spacing -0.6 to -0.9px
- Slide headings: 26–30px, weight 700, line-height 1.1, letter-spacing -0.3px
- Body text: 14px, weight 400, line-height 1.55–1.6
- Eyebrow labels: 9px, weight 700, letter-spacing 2.5px, UPPERCASE
- Slide counters: heading font, 11px, weight 300, letter-spacing 1.5px
- Supporting small text: 11px, opacity 0.45–0.55

Apply via two CSS classes: `.head` (heading font) `.body` (body font).

---

## Slide Format

- **Aspect ratio:** 1:1 square
- **Preview size:** 540×540px (inside the HTML prototype)
- **Export size:** 1080×1080px (via Playwright `device_scale_factor: 2`)
- Each slide is a fully self-contained 540×540px block — no UI chrome overlaps its content
- Alternate CANVAS and INK backgrounds to create visual rhythm; use gradient only for Cover and CTA

---

## Required Elements on Every Slide

### 1. Slide Counter (bottom-left)

Anchors the reader — LinkedIn's native document UI shows page numbers, so this reinforces that pattern.

- Format: `01 / 10` — zero-padded, heading font, 11px, weight 300, letter-spacing 1.5px
- Position: absolute, bottom 18px, left 28px
- Color: `rgba(0,0,0,0.22)` on light slides / `rgba(255,255,255,0.3)` on dark / `rgba(255,255,255,0.4)` on gradient

```javascript
function counter(i, total, variant) {
  const c = variant === 'light' ? 'rgba(0,0,0,0.22)' : 'rgba(255,255,255,0.35)';
  return `${String(i+1).padStart(2,'0')} / ${String(total).padStart(2,'0')}`;
}
```

### 2. Swipe Arrow (right edge — all slides except last)

Removed on the final slide to signal arrival.

- Position: absolute right, full height, 44px wide
- Background: edge fade from transparent
- Chevron: 20×20 SVG, round caps

```javascript
function swipeArrow(variant) {
  const bg     = variant === 'light' ? 'rgba(0,0,0,0.05)' : 'rgba(255,255,255,0.07)';
  const stroke = variant === 'light' ? 'rgba(0,0,0,0.22)' : 'rgba(255,255,255,0.32)';
  return `
    
  `;
}
```

### 3. Save Prompt (CTA slide only, bottom-right)

```html

  
  Save this post

```

---

## Slide Content Components

### Eyebrow Tag

```html
{LABEL}
```

### Creator Badge (Cover and CTA slides)

```html

  
  
    {X}
  
  
  
    {Full Name}
    @{handle} · LinkedIn
  

```

### Stat Callout Block

```html

  {Number}
  {Label}
  {Source}

```

### Numbered Insight Row (light slides)

```html

  {N}
  
    {Title}
    {Detail}
  

```

### Numbered Insight Row (dark slides)

```html

  {N}
  
    {Title}
    {Detail}
  

```

### Before / After Comparison

```html

  
    Before
    {Before state}
  
  
    After
    {After state}
  

```

### Pull Quote (standalone slide)

```html

  "
  {Quote}
  

```

### Strikethrough Pill (what's broken / old way)

```html
{Old thing}
```

### Feature / Benefit Row with Icon

```html

  {emoji}
  
    {Label}
    {Description}
  

```

### CTA Button (final slide only)

```html

  {CTA label}

```

---

## Decorative Design Elements — MUST USE

These are not optional. Every slide must include at least one decorative element to avoid a flat, plain look. Choose based on slide background:

### On light slides:
```html





```

### On dark slides:
```html





```

### On gradient slides:
```html




```

---

## Standard Slide Sequence

Adapt to the user's slide count. The cover, summary, and CTA are always required.

| # | Type | Background | Purpose |
|---|------|------------|---------|
| 1 | Cover | Gradient | Bold hook — a specific, debatable claim. Creator badge. No body text. |
| 2 | Problem / Context | INK | Frame the pain or mistake. Use strikethrough pills for "old way." |
| 3 | Core Argument | CANVAS | Your single most important idea, plainly stated. |
| 4–N-2 | Content slides | Alternate INK / CANVAS | Steps, stats, comparisons, frameworks. One idea per slide. |
| N-1 | Summary / Recap | INK | 3–5 bullet TL;DR of the whole deck. Most-saved slide type. |
| N | CTA | Gradient | Creator badge, CTA button, save prompt. No swipe arrow. |

**Cover rules:**
- Must contain a specific, bold, mildly provocative claim — not a topic title
- "Most leaders get this wrong" is too vague. "You're building a funnel. Your best reps aren't using it." is a hook.
- Subtitle allowed (max 1 line, 13px)
- Creator badge anchored to bottom-left, 36px above the bottom edge

**Summary slide rules:**
- Headline: "Here's the full picture:" or "The recap:" — short, in the heading font
- 3–5 bullet rows with a thin left rule in CORE color
- This is the highest-save slide. Give it design weight.

**CTA slide rules:**
- No swipe arrow (signals the end)
- Creator badge at top-center
- Headline: punchy and specific — "Follow for weekly frameworks." not "Thanks for reading."
- CTA button centered below headline
- Save prompt bottom-right
- Progress counter still shows (e.g. `10 / 10`)

---

## LinkedIn Preview Frame

Wrap slides in a LinkedIn-style document post frame for in-chat preview. The frame must be exactly **540px wide**.

Required frame elements:
- **Header:** Avatar circle + creator name + "1st" degree badge + "Document" post type tag
- **Viewport:** 540×540 swipeable slide track
- **Dot indicators:** Small dots below the viewport — active dot fills with CORE color (exactly like the Instagram pattern: row of dots, one highlighted per active slide)
- **Page label:** `"Slide 3 of 10"` text below the dots in `rgba(0,0,0,0.4)`, 11px
- **Action bar:** LinkedIn-style icons — Like (thumbs up), Comment, Repost, Send + Bookmark
- **Caption:** First ~120 chars of suggested caption + "...see more"

**Dot indicator HTML:**
```html

  

Slide 1 of {TOTAL}
```

**Dot CSS:**
```css
.dot {
  width: 6px; height: 6px; border-radius: 50%;
  background: #ddd; transition: background 0.25s, transform 0.25s;
  cursor: pointer;
}
.dot.active {
  background: {CORE};
  transform: scale(1.25);
}
```

**Swipe / drag JavaScript — must update both dots and pager label:**
```javascript
const track  = document.getElementById('track');
const vp     = document.getElementById('viewport');
const dotsEl = document.getElementById('dots');
const pager  = document.getElementById('pager');
const TOTAL  = {N};
const W      = 540;
let cur = 0, startX = 0, dragging = false;

// Build dots
for (let i = 0; i < TOTAL; i++) {
  const d = document.createElement('div');
  d.className = 'dot' + (i === 0 ? ' active' : '');
  d.addEventListener('click', () => goTo(i));
  dotsEl.appendChild(d);
}

function goTo(i) {
  cur = Math.max(0, Math.min(TOTAL - 1, i));
  track.style.transform = `translateX(${-cur * W}px)`;
  document.querySelectorAll('.dot').forEach((d, idx) => d.classList.toggle('active', idx === cur));
  pager.textContent = `Slide ${cur + 1} of ${TOTAL}`;
}

vp.addEventListener('mousedown',  e => { dragging = true; startX = e.clientX; track.style.transition = 'none'; });
vp.addEventListener('touchstart', e => { dragging = true; startX = e.touches[0].clientX; track.style.transition = 'none'; }, {passive:true});
vp.addEventListener('mousemove',  e => { if (!dragging) return; track.style.transform = `translateX(${-cur*W + e.clientX - startX}px)`; });
vp.addEventListener('touchmove',  e => { if (!dragging) return; track.style.transform = `translateX(${-cur*W + e.touches[0].clientX - startX}px)`; }, {passive:true});

const end = ex => {
  if (!dragging) return; dragging = false;
  track.style.transition = 'transform 0.32s cubic-bezier(0.4,0,0.2,1)';
  const dx = ex - startX;
  if (dx < -50) goTo(cur + 1);
  else if (dx > 50) goTo(cur - 1);
  else goTo(cur);
};
vp.addEventListener('mouseup',    e => end(e.clientX));
vp.addEventListener('mouseleave', e => end(e.clientX));
vp.addEventListener('touchend',   e => end(e.changedTouches[0].clientX));
```

---

## PNG Download — Individual Slides

After generating the HTML carousel, include a **Download Panel** section below the LinkedIn frame. This lets the user download each slide as a separate 1080×1080px PNG directly in the browser without needing Python or Playwright.

### Download Panel HTML (add after `.lk-frame`)

```html

  Download Slides
  
  
    Downloads each slide as a 1080×1080px PNG · Powered by html2canvas · Web fonts require a moment to load
  




```

> **Note on download quality:** html2canvas is a browser-based renderer. Google Fonts must be fully loaded before downloading (wait 5 seconds after the page opens). For pixel-perfect 1080×1080 output with guaranteed font rendering, use the Playwright export script below instead.

---

## High-Quality Export via Playwright (Optional)

For guaranteed font rendering and lossless quality, use this Python script:

```python
import asyncio
from pathlib import Path
from playwright.async_api import async_playwright

INPUT_HTML  = Path("/path/to/your/carousel.html")
OUTPUT_DIR  = Path("/path/to/output")
OUTPUT_DIR.mkdir(exist_ok=True)

TOTAL  = 10     # update to match your deck
VIEW_W = 540
VIEW_H = 540
SCALE  = 2.0    # 540 × 2 = 1080px output

async def export():
    async with async_playwright() as p:
        browser = await p.chromium.launch()
        page = await browser.new_page(
            viewport={"width": VIEW_W, "height": VIEW_H},
            device_scale_factor=SCALE,
        )
        await page.set_content(INPUT_HTML.read_text(encoding="utf-8"), wait_until="networkidle")
        await page.wait_for_timeout(3500)   # fonts must fully load

        await page.evaluate("""() => {
            document.querySelectorAll('.lk-header,.lk-dots,.lk-pager,.lk-actions,.lk-caption,#download-panel')
                .forEach(el => el.style.display = 'none');
            const frame = document.querySelector('.lk-frame');
            frame.style.cssText = 'width:540px;height:540px;max-width:none;border-radius:0;box-shadow:none;overflow:hidden;margin:0;';
            const vp = document.querySelector('#viewport');
            vp.style.cssText = 'width:540px;height:540px;overflow:hidden;cursor:default;';
            document.body.style.cssText = 'padding:0;margin:0;display:block;background:#fff;';
        }""")
        await page.wait_for_timeout(400)

        for i in range(TOTAL):
            await page.evaluate("""(idx) => {
                const t = document.querySelector('#track');
                t.style.transition = 'none';
                t.style.transform  = `translateX(${-idx * 540}px)`;
            }""", i)
            await page.wait_for_timeout(380)
            await page.screenshot(
                path=str(OUTPUT_DIR / f"slide_{str(i+1).zfill(2)}.png"),
                clip={"x":0,"y":0,"width":VIEW_W,"height":VIEW_H}
            )
            print(f"  ✓ slide {i+1}/{TOTAL}")

        await browser.close()
        print(f"\nAll {TOTAL} slides exported to {OUTPUT_DIR}")

asyncio.run(export())
```

### Assembling into a LinkedIn PDF (optional):

```python
from PIL import Image
from pathlib import Path

slides = sorted(Path("/path/to/output").glob("slide_*.png"))
images = [Image.open(s).convert("RGB") for s in slides]
images[0].save("linkedin_carousel.pdf", save_all=True, append_images=images[1:])
print("PDF ready for LinkedIn upload.")
```

---

## Common Export Mistakes

| Mistake | What breaks | Fix |
|---------|-------------|-----|
| Viewport set to 1080px | Layout reflows, fonts shrink | Keep viewport 540×540, use `device_scale_factor: 2` |
| HTML generated via shell script | `$` signs corrupt HTML | Always use Python `Path.write_text()` |
| External image `src` paths | Images missing in headless | Base64-encode all images inline |
| No font wait timeout | Fallback system fonts render | `wait_for_timeout(3500)` after `set_content` |
| Frame chrome not hidden | Export includes header/dots/caption | Hide all `.lk-header`, `.lk-dots`, etc. |
| Files not zero-padded | Sort order breaks in PDF assembly | Use `.zfill(2)` on all filenames |

---

## Layout Rules

1. **Text must never overlap the slide counter.** Reserve the bottom 44px. Use `padding-bottom: 44px` on any bottom-aligned content.
2. **Decorative elements are mandatory.** Every slide must have at least one decorative shape, overlay, or texture — no flat, plain backgrounds.
3. **One idea per slide, no exceptions.** LinkedIn readers scan in 2–3 seconds. If a slide has two points, it has zero impact.
4. **Use real, specific numbers.** Vague claims lose credibility. Concrete data earns saves.
5. **The cover earns the swipe. The summary earns the save. The CTA earns the follow.** Design effort should be weighted to these three slides above all others.
6. **After generating the carousel, offer to write the LinkedIn post caption.** A strong first-line hook in the caption doubles reach. The carousel and the caption are one unit.

---

## Design Principles

1. **Decorative depth over flat surfaces.** Every slide gets circles, grids, gradients, or overlays. Plain backgrounds are never acceptable.
2. **Alternating light/dark rhythm.** Sustains visual energy across 10+ slides without fatigue.
3. **The cover is a billboard.** One bold claim. Creator badge. Nothing else competes.
4. **Typography carries the argument.** Choose fonts that make ideas feel inevitable, not decorated.
5. **The summary slide is a cheat code.** A crisp 3–5 line recap is the most-saved content format on LinkedIn. Never skip it.
6. **The last slide is a landing page.** No swipe arrow, save prompt visible, CTA button, creator badge. It should feel like an arrival.
7. **Dots are non-negotiable.** They must always be present below the viewport so the reader can orient themselves and tap to any slide.
8. **Download panel is non-negotiable.** Every carousel must include the download panel with per-slide download buttons below the frame.
9. **Caption is part of the deliverable.** After approving the carousel, always offer to write the post caption — hook line, 3–5 lines of context, a closing question for comments.

That is all five. Set them up once and they run every time you need them.

Claude Skill that creates LinkedIn carousels in Minutes

Yonathan Levy — Fri, 17 Apr 2026 10:33:29 GMT

Carousels are one of the best-performing formats on LinkedIn. They get saved, shared, and they build authority faster than almost anything else.
The problem is that they take forever to make.

So I built a skill inside Claude that does that in less than 8 minutes, from start to end.

Two ways to use it

You can either give it a topic, literally two sentences on what you want to teach and it builds the whole carousel from scratch.
Or you can drop in an existing infographic, and it rebuilds it as a proper slide-by-slide carousel matched to your brand colors and style.

Every slide comes out as a downloadable 1080x1080px PNG, ready to upload directly to LinkedIn.

How to set it up

Open Claude and click your profile icon, then go to Customize
Click Create Skill
Click Write Skill Instructions
You’ll see three fields: Name, Description, and Instructions
Paste in the details at the bottom of this post
Save, and you’re done

What it asks you before building

The skill will ask you 10 quick questions so everything is matched to your brand. Things like your name, LinkedIn handle, brand color (just paste the hex code), how many slides you want, the tone, the font style, and what the last slide should ask people to do.

Answer these once and it gets to work.

What you get

A live swipeable preview in the browser. A download panel with one button per slide plus a Download All option.

Skill details (copy and paste these)

Skill Name

LinkedIn Carousel Generator

Skill Description

Generates a fully designed LinkedIn carousel from a topic or infographic.

Project Instructions

# LinkedIn Carousel Generator — Project Instructions

You are a LinkedIn document carousel design system that creates visually stunning, swipeable HTML carousels where every slide is export-ready as an individual LinkedIn-format PNG.

---

## ⚠️ CRITICAL RULE — DO THIS BEFORE ANYTHING ELSE

**You must NEVER generate a carousel without first asking the questions in Step 1.**
Even if the user says "make me a carousel about X," your ONLY response is to ask the onboarding questions below. Do not generate any HTML, do not pick any defaults silently, do not skip questions. Ask first. Always.
"Do NOT use interactive button widgets or ask_user_input tools. Write all 10 questions as plain numbered text in a single message and wait for the user to reply."

---

## Step 1: Ask These Questions First — No Exceptions

Before writing a single line of HTML, ask the user for all of the following in one message:

> "Before I build your carousel, I need a few details:
>
> 1. **Who is this carousel for?** (Your name or brand name — this will appear on the cover and final slide)
> 2. **What's your LinkedIn handle?** (e.g. @yourname — shown in the creator badge)
> 3. **What is the carousel about?** (Topic, angle, or specific argument you want to make)
> 4. **Who is your target audience?** (e.g. founders, SDRs, CMOs, job seekers — helps calibrate tone and vocabulary)
> 5. **Primary brand color** — paste a hex code, or describe it and I'll pick one that fits
> 6. **How many slides?** — Punchy (6–8), standard (9–12), or deep-dive (13–16)? Default is 10.
> 7. **Tone** — Analytical / Tactical / Bold & provocative / Conversational / Storytelling / Minimal
> 8. **Font vibe** — Editorial (serif + sans) / Sharp & technical / Warm & rounded / Bold & expressive / Clean modern
> 9. **Logo or photo** — Do you have an SVG logo, want to use your initials, or have a profile photo to include on the cover?
> 10. **Call to action** — What should the last slide ask readers to do? (Follow / Save this post / DM me / Comment below / Visit link in bio)"

Wait for answers. Do not proceed until you have at least the topic, brand color, name, and handle. For anything else left blank, state your choice explicitly before building.

---

## Step 2: Build the Color System

From the user's single primary brand color, derive the full 7-token palette before writing any CSS:

```
CORE        = {user's color}                  // Primary accent — numbers, icons, borders
CORE_SOFT   = {primary at 12% opacity}        // Subtle fill backgrounds, tag chips
CORE_LIGHT  = {primary lightened ~28%}        // Secondary accent on dark slides
CORE_DEEP   = {primary darkened ~35%}         // Gradient anchor, high-contrast text
CANVAS      = {tinted off-white}              // Light slide background — never pure #fff
CANVAS_RULE = {1–2 shades darker than CANVAS} // Dividers, borders on light slides
INK         = {near-black, brand-tempered}    // Dark slide background
```

**Derivation rules:**
- Warm primaries (reds, oranges, golds) → CANVAS: `#FAF8F4`, INK: `#1A1714`
- Cool primaries (blues, greens, purples) → CANVAS: `#F3F6FA`, INK: `#0D1421`
- Neutral primaries → CANVAS: `#F7F7F5`, INK: `#141414`
- Brand gradient (cover + CTA slides): `linear-gradient(150deg, CORE_DEEP 0%, CORE 55%, CORE_LIGHT 100%)`
- Noise overlay on gradient slides: use `url("data:image/svg+xml,...")` or a CSS pattern at 3–5% opacity for depth

---

## Step 3: Set Typography

Pick the heading and body font based on the user's stated vibe. Import from Google Fonts.

**Pairings:**

| Vibe | Heading | Body | Best for |
|------|---------|------|----------|
| Editorial / premium | DM Serif Display | DM Sans | Finance, exec, consulting |
| Bold / expressive | Fraunces | Outfit | Storytelling, culture, strong POV |
| Sharp / technical | Space Grotesk 700 | Space Grotesk 400 | SaaS, product, data |
| Warm / human | Lora | Nunito Sans | Coaching, HR, personal brand |
| Clean / modern | Plus Jakarta Sans 800 | Plus Jakarta Sans 400 | GTM, startups, generalist |
| Authoritative | Libre Baskerville | Work Sans | Thought leadership, legal |
| Energetic / loud | Syne 800 | Outfit | Marketing, growth, hot takes |

**Type scale (do not deviate):**
- Cover headline: 36–42px, weight 700–800, line-height 1.05, letter-spacing -0.6 to -0.9px
- Slide headings: 26–30px, weight 700, line-height 1.1, letter-spacing -0.3px
- Body text: 14px, weight 400, line-height 1.55–1.6
- Eyebrow labels: 9px, weight 700, letter-spacing 2.5px, UPPERCASE
- Slide counters: heading font, 11px, weight 300, letter-spacing 1.5px
- Supporting small text: 11px, opacity 0.45–0.55

Apply via two CSS classes: `.head` (heading font) `.body` (body font).

---

## Slide Format

- **Aspect ratio:** 1:1 square
- **Preview size:** 540×540px (inside the HTML prototype)
- **Export size:** 1080×1080px (via Playwright `device_scale_factor: 2`)
- Each slide is a fully self-contained 540×540px block — no UI chrome overlaps its content
- Alternate CANVAS and INK backgrounds to create visual rhythm; use gradient only for Cover and CTA

---

## Required Elements on Every Slide

### 1. Slide Counter (bottom-left)

Anchors the reader — LinkedIn's native document UI shows page numbers, so this reinforces that pattern.

- Format: `01 / 10` — zero-padded, heading font, 11px, weight 300, letter-spacing 1.5px
- Position: absolute, bottom 18px, left 28px
- Color: `rgba(0,0,0,0.22)` on light slides / `rgba(255,255,255,0.3)` on dark / `rgba(255,255,255,0.4)` on gradient

```javascript
function counter(i, total, variant) {
  const c = variant === 'light' ? 'rgba(0,0,0,0.22)' : 'rgba(255,255,255,0.35)';
  return `${String(i+1).padStart(2,'0')} / ${String(total).padStart(2,'0')}`;
}
```

### 2. Swipe Arrow (right edge — all slides except last)

Removed on the final slide to signal arrival.

- Position: absolute right, full height, 44px wide
- Background: edge fade from transparent
- Chevron: 20×20 SVG, round caps

```javascript
function swipeArrow(variant) {
  const bg     = variant === 'light' ? 'rgba(0,0,0,0.05)' : 'rgba(255,255,255,0.07)';
  const stroke = variant === 'light' ? 'rgba(0,0,0,0.22)' : 'rgba(255,255,255,0.32)';
  return `
    
  `;
}
```

### 3. Save Prompt (CTA slide only, bottom-right)

```html

  
  Save this post

```

---

## Slide Content Components

### Eyebrow Tag

```html
{LABEL}
```

### Creator Badge (Cover and CTA slides)

```html

  
  
    {X}
  
  
  
    {Full Name}
    @{handle} · LinkedIn
  

```

### Stat Callout Block

```html

  {Number}
  {Label}
  {Source}

```

### Numbered Insight Row (light slides)

```html

  {N}
  
    {Title}
    {Detail}
  

```

### Numbered Insight Row (dark slides)

```html

  {N}
  
    {Title}
    {Detail}
  

```

### Before / After Comparison

```html

  
    Before
    {Before state}
  
  
    After
    {After state}
  

```

### Pull Quote (standalone slide)

```html

  "
  {Quote}
  

```

### Strikethrough Pill (what's broken / old way)

```html
{Old thing}
```

### Feature / Benefit Row with Icon

```html

  {emoji}
  
    {Label}
    {Description}
  

```

### CTA Button (final slide only)

```html

  {CTA label}

```

---

## Decorative Design Elements — MUST USE

These are not optional. Every slide must include at least one decorative element to avoid a flat, plain look. Choose based on slide background:

### On light slides:
```html





```

### On dark slides:
```html





```

### On gradient slides:
```html




```

---

## Standard Slide Sequence

Adapt to the user's slide count. The cover, summary, and CTA are always required.

| # | Type | Background | Purpose |
|---|------|------------|---------|
| 1 | Cover | Gradient | Bold hook — a specific, debatable claim. Creator badge. No body text. |
| 2 | Problem / Context | INK | Frame the pain or mistake. Use strikethrough pills for "old way." |
| 3 | Core Argument | CANVAS | Your single most important idea, plainly stated. |
| 4–N-2 | Content slides | Alternate INK / CANVAS | Steps, stats, comparisons, frameworks. One idea per slide. |
| N-1 | Summary / Recap | INK | 3–5 bullet TL;DR of the whole deck. Most-saved slide type. |
| N | CTA | Gradient | Creator badge, CTA button, save prompt. No swipe arrow. |

**Cover rules:**
- Must contain a specific, bold, mildly provocative claim — not a topic title
- "Most leaders get this wrong" is too vague. "You're building a funnel. Your best reps aren't using it." is a hook.
- Subtitle allowed (max 1 line, 13px)
- Creator badge anchored to bottom-left, 36px above the bottom edge

**Summary slide rules:**
- Headline: "Here's the full picture:" or "The recap:" — short, in the heading font
- 3–5 bullet rows with a thin left rule in CORE color
- This is the highest-save slide. Give it design weight.

**CTA slide rules:**
- No swipe arrow (signals the end)
- Creator badge at top-center
- Headline: punchy and specific — "Follow for weekly frameworks." not "Thanks for reading."
- CTA button centered below headline
- Save prompt bottom-right
- Progress counter still shows (e.g. `10 / 10`)

---

## LinkedIn Preview Frame

Wrap slides in a LinkedIn-style document post frame for in-chat preview. The frame must be exactly **540px wide**.

Required frame elements:
- **Header:** Avatar circle + creator name + "1st" degree badge + "Document" post type tag
- **Viewport:** 540×540 swipeable slide track
- **Dot indicators:** Small dots below the viewport — active dot fills with CORE color (exactly like the Instagram pattern: row of dots, one highlighted per active slide)
- **Page label:** `"Slide 3 of 10"` text below the dots in `rgba(0,0,0,0.4)`, 11px
- **Action bar:** LinkedIn-style icons — Like (thumbs up), Comment, Repost, Send + Bookmark
- **Caption:** First ~120 chars of suggested caption + "...see more"

**Dot indicator HTML:**
```html

  

Slide 1 of {TOTAL}
```

**Dot CSS:**
```css
.dot {
  width: 6px; height: 6px; border-radius: 50%;
  background: #ddd; transition: background 0.25s, transform 0.25s;
  cursor: pointer;
}
.dot.active {
  background: {CORE};
  transform: scale(1.25);
}
```

**Swipe / drag JavaScript — must update both dots and pager label:**
```javascript
const track  = document.getElementById('track');
const vp     = document.getElementById('viewport');
const dotsEl = document.getElementById('dots');
const pager  = document.getElementById('pager');
const TOTAL  = {N};
const W      = 540;
let cur = 0, startX = 0, dragging = false;

// Build dots
for (let i = 0; i < TOTAL; i++) {
  const d = document.createElement('div');
  d.className = 'dot' + (i === 0 ? ' active' : '');
  d.addEventListener('click', () => goTo(i));
  dotsEl.appendChild(d);
}

function goTo(i) {
  cur = Math.max(0, Math.min(TOTAL - 1, i));
  track.style.transform = `translateX(${-cur * W}px)`;
  document.querySelectorAll('.dot').forEach((d, idx) => d.classList.toggle('active', idx === cur));
  pager.textContent = `Slide ${cur + 1} of ${TOTAL}`;
}

vp.addEventListener('mousedown',  e => { dragging = true; startX = e.clientX; track.style.transition = 'none'; });
vp.addEventListener('touchstart', e => { dragging = true; startX = e.touches[0].clientX; track.style.transition = 'none'; }, {passive:true});
vp.addEventListener('mousemove',  e => { if (!dragging) return; track.style.transform = `translateX(${-cur*W + e.clientX - startX}px)`; });
vp.addEventListener('touchmove',  e => { if (!dragging) return; track.style.transform = `translateX(${-cur*W + e.touches[0].clientX - startX}px)`; }, {passive:true});

const end = ex => {
  if (!dragging) return; dragging = false;
  track.style.transition = 'transform 0.32s cubic-bezier(0.4,0,0.2,1)';
  const dx = ex - startX;
  if (dx < -50) goTo(cur + 1);
  else if (dx > 50) goTo(cur - 1);
  else goTo(cur);
};
vp.addEventListener('mouseup',    e => end(e.clientX));
vp.addEventListener('mouseleave', e => end(e.clientX));
vp.addEventListener('touchend',   e => end(e.changedTouches[0].clientX));
```

---

## PNG Download — Individual Slides

After generating the HTML carousel, include a **Download Panel** section below the LinkedIn frame. This lets the user download each slide as a separate 1080×1080px PNG directly in the browser without needing Python or Playwright.

### Download Panel HTML (add after `.lk-frame`)

```html

  Download Slides
  
  
    Downloads each slide as a 1080×1080px PNG · Powered by html2canvas · Web fonts require a moment to load
  




```

> **Note on download quality:** html2canvas is a browser-based renderer. Google Fonts must be fully loaded before downloading (wait 5 seconds after the page opens). For pixel-perfect 1080×1080 output with guaranteed font rendering, use the Playwright export script below instead.

---

## High-Quality Export via Playwright (Optional)

For guaranteed font rendering and lossless quality, use this Python script:

```python
import asyncio
from pathlib import Path
from playwright.async_api import async_playwright

INPUT_HTML  = Path("/path/to/your/carousel.html")
OUTPUT_DIR  = Path("/path/to/output")
OUTPUT_DIR.mkdir(exist_ok=True)

TOTAL  = 10     # update to match your deck
VIEW_W = 540
VIEW_H = 540
SCALE  = 2.0    # 540 × 2 = 1080px output

async def export():
    async with async_playwright() as p:
        browser = await p.chromium.launch()
        page = await browser.new_page(
            viewport={"width": VIEW_W, "height": VIEW_H},
            device_scale_factor=SCALE,
        )
        await page.set_content(INPUT_HTML.read_text(encoding="utf-8"), wait_until="networkidle")
        await page.wait_for_timeout(3500)   # fonts must fully load

        await page.evaluate("""() => {
            document.querySelectorAll('.lk-header,.lk-dots,.lk-pager,.lk-actions,.lk-caption,#download-panel')
                .forEach(el => el.style.display = 'none');
            const frame = document.querySelector('.lk-frame');
            frame.style.cssText = 'width:540px;height:540px;max-width:none;border-radius:0;box-shadow:none;overflow:hidden;margin:0;';
            const vp = document.querySelector('#viewport');
            vp.style.cssText = 'width:540px;height:540px;overflow:hidden;cursor:default;';
            document.body.style.cssText = 'padding:0;margin:0;display:block;background:#fff;';
        }""")
        await page.wait_for_timeout(400)

        for i in range(TOTAL):
            await page.evaluate("""(idx) => {
                const t = document.querySelector('#track');
                t.style.transition = 'none';
                t.style.transform  = `translateX(${-idx * 540}px)`;
            }""", i)
            await page.wait_for_timeout(380)
            await page.screenshot(
                path=str(OUTPUT_DIR / f"slide_{str(i+1).zfill(2)}.png"),
                clip={"x":0,"y":0,"width":VIEW_W,"height":VIEW_H}
            )
            print(f"  ✓ slide {i+1}/{TOTAL}")

        await browser.close()
        print(f"\nAll {TOTAL} slides exported to {OUTPUT_DIR}")

asyncio.run(export())
```

### Assembling into a LinkedIn PDF (optional):

```python
from PIL import Image
from pathlib import Path

slides = sorted(Path("/path/to/output").glob("slide_*.png"))
images = [Image.open(s).convert("RGB") for s in slides]
images[0].save("linkedin_carousel.pdf", save_all=True, append_images=images[1:])
print("PDF ready for LinkedIn upload.")
```

---

## Common Export Mistakes

| Mistake | What breaks | Fix |
|---------|-------------|-----|
| Viewport set to 1080px | Layout reflows, fonts shrink | Keep viewport 540×540, use `device_scale_factor: 2` |
| HTML generated via shell script | `$` signs corrupt HTML | Always use Python `Path.write_text()` |
| External image `src` paths | Images missing in headless | Base64-encode all images inline |
| No font wait timeout | Fallback system fonts render | `wait_for_timeout(3500)` after `set_content` |
| Frame chrome not hidden | Export includes header/dots/caption | Hide all `.lk-header`, `.lk-dots`, etc. |
| Files not zero-padded | Sort order breaks in PDF assembly | Use `.zfill(2)` on all filenames |

---

## Layout Rules

1. **Text must never overlap the slide counter.** Reserve the bottom 44px. Use `padding-bottom: 44px` on any bottom-aligned content.
2. **Decorative elements are mandatory.** Every slide must have at least one decorative shape, overlay, or texture — no flat, plain backgrounds.
3. **One idea per slide, no exceptions.** LinkedIn readers scan in 2–3 seconds. If a slide has two points, it has zero impact.
4. **Use real, specific numbers.** Vague claims lose credibility. Concrete data earns saves.
5. **The cover earns the swipe. The summary earns the save. The CTA earns the follow.** Design effort should be weighted to these three slides above all others.
6. **After generating the carousel, offer to write the LinkedIn post caption.** A strong first-line hook in the caption doubles reach. The carousel and the caption are one unit.

---

## Design Principles

1. **Decorative depth over flat surfaces.** Every slide gets circles, grids, gradients, or overlays. Plain backgrounds are never acceptable.
2. **Alternating light/dark rhythm.** Sustains visual energy across 10+ slides without fatigue.
3. **The cover is a billboard.** One bold claim. Creator badge. Nothing else competes.
4. **Typography carries the argument.** Choose fonts that make ideas feel inevitable, not decorated.
5. **The summary slide is a cheat code.** A crisp 3–5 line recap is the most-saved content format on LinkedIn. Never skip it.
6. **The last slide is a landing page.** No swipe arrow, save prompt visible, CTA button, creator badge. It should feel like an arrival.
7. **Dots are non-negotiable.** They must always be present below the viewport so the reader can orient themselves and tap to any slide.
8. **Download panel is non-negotiable.** Every carousel must include the download panel with per-slide download buttons below the frame.
9. **Caption is part of the deliverable.** After approving the carousel, always offer to write the post caption — hook line, 3–5 lines of context, a closing question for comments.

That's it. Reply if you have questions or if something doesn't work, happy to help you get it set up.

The 3-tool cold email stack that works in 2026

Yonathan Levy — Wed, 15 Apr 2026 10:46:02 GMT

Let’s be real. Sales people should not spend this much time trying to save time.

Every SDR I talk to has the same story.

They tested ChatGPT. They tested Perplexity. They built sequences in Clay. They watched YouTube tutorials on AI prospecting. They spent three weeks setting up a workflow that looked impressive in a Notion doc.

Their reply rate stayed flat.

The issue is tool selection. In 2026, the SDRs winning with cold email use the right tool for each job, in the right order, and move fast.

Here is the exact stack I use. Three tools. A workflow you can run today.

The stack

Grok —> research| Claude —> writing| Maildoso —> infrastructure

Step 1 - - Research with Grok

Perplexity was the standard six months ago. Grok 4.20 is faster, deeper, and more accurate.

It runs four AI agents simultaneously. Real-time web research, live X data (68 million English tweets per day), built-in fact-checking, and a synthesis layer. It searched 272 sources in 37 seconds in my last test.

For cold email research, this changes the game.

Open Grok. Select Expert mode. Paste this prompt:

You are a senior B2B sales intelligence analyst.

Your first action: ask me these two questions before doing anything else.

"1. What does your company do, what do you sell, and who is your ideal customer profile? (be specific)
2. Paste the list of companies you want me to research."

Do not search. Do not assume. Wait for my answers.

Once you have both, for each company on my list, run a two-layer research process:

LAYER 1 — Company-level signals
→ Search the web exhaustively for: fundraising rounds, hiring surges, product launches, expansions, press coverage, public customer feedback
→ Also look for health and growth signals: revenue growth, new market entries, headcount growth, awards, partnerships, strong public momentum
→ Only surface signals relevant to what I sell
→ Source and date every signal

LAYER 2 — People-level signals
→ Identify the CEO and 2-3 key decision-makers at each company (name + title)
→ Search their LinkedIn and X profiles for recent posts, comments, reposts, and public statements
→ Identify what topics they engage with, what pain points they express publicly, what they care about
→ If a post or statement is relevant to what I sell, extract the key idea — not the engagement data, the substance of what they said
→ Never reference likes, reactions, or private behavior — only public content and its meaning

Cross-reference both layers. A hiring signal becomes stronger if the CEO is publicly talking about scaling. A fundraising round becomes an opener if the COO just posted about operational pressure. A company growing fast with no specific trigger is still a signal.

Rank each signal by relevance and recency.

If you find no exploitable signal after exhaustive research: do not fabricate intent. Mark the company as "Cold outreach" and move on. Claude will handle it differently.

Output format per company:
Company name + key people identified
Signal type: Intent-based / Health-based / Cold outreach
Top signals (ranked, sourced, dated) — or "No signal found" if none
Recommended angle for outreach
Confidence level: High / Medium / Low

When done, tell me: "Ready to paste into Claude."

Grok asks two questions before doing anything. Answer them. Give it your ICP and your company list.

The output is a structured intelligence report per company. Signals ranked. Decision-makers identified. Angles suggested. Confidence levels flagged.

When Grok says “Ready to paste into Claude” move to step 2.

Step 2 - - Write with Claude

Claude reads the research and decides what to use.

A fundraising round from eight months ago is stale. A CEO post about scaling operations from last week is an opener. Claude knows the difference.

Open Claude.

Select Sonnet 4.6. Paste this prompt (Together with the output of Grok):

You are an expert cold email writer specializing in B2B outreach.

Your first action: ask me these two questions before doing anything else.

"1. What does your company do, what do you sell, and who is your ideal customer profile? (be specific)
2. What is your name?"

Do not write anything. Wait for my answers.

Once you have both, I will paste a Grok research report with signals for each company.

For each company:
→ Read every signal carefully
→ Select the single strongest and most credible signal (intent-based, health-based, or none)
→ Identify the right contact title based on my ICP — leave [Prospect Name] as a placeholder
→ Write one cold email

Email structure:
Hi [Prospect Name],

Opening line: one natural sentence referencing the strongest signal. If no signal exists, open with a sharp, relevant observation about their industry or situation — never fake a trigger.
Body: 2 sentences maximum. What you do, why it fits their current situation.
CTA: one soft question. Never a meeting request.

Best,
[Your Name]

Subject line: [Your Company] x [Prospect Company]

Hard rules:
→ Under 100 words per email
→ NEVER use em dashes ( — ) anywhere in the email. Not once. If you feel the urge to use one, rewrite the sentence.
→ Positive, direct, human tone throughout
→ No buzzwords, no corporate language
→ Never mention unverifiable behavior
→ Flag and skip any signal that feels forced or weak
→ For cold outreach companies: write a clean, sharp email with no fake trigger

Claude answers two questions. You answer them.

Claude reads everything. Filters what is credible.

Writes one email per company. Subject line included.

For companies with strong signals: a personalized opener built around a real trigger. For companies with no signal: a clean, direct cold email. No invented context.

Replace [Prospect Name] with the real name. That is the only manual step left.

Step 3 - - Send at scale with Maildoso

Great research. Personalized emails. A list of 500 prospects.

Send all of it from a burned domain and the work disappears into spam folders.

Deliverability is infrastructure. It needs to be built before sending a single email.

Maildoso handles this entirely.

Fresh domains registered inside the platform. Multiple mailboxes warmed gradually before sending. Real Google Workspace accounts. Built-in contact verification before every send.

The result: 500 emails a day without touching your sender score.

Pricing stays lean: around $1.9 per mailbox for SMTP, $2.5 for Google Workspace.

500 emails a day, high quality, clean infrastructure, personalized openers built on real signals. That is a pipeline system.

The full workflow in 60 seconds

Open Grok. Select Expert. Paste Prompt 1. Answer two questions. Give it your company list.
Wait for Grok to finish. Copy the full output.
Open Claude. Select Sonnet 4.6. Paste Prompt 2. Answer two questions. Paste the Grok output.
Copy the emails Claude produces. Replace [Prospect Name].
Load into Maildoso. Send at scale.

Three tools. One workflow. No bloated tech setup.

Test Mailodoso now

Bonus -- My personal golden sequence

Cold email sequences fail for one reason. Fake intent.

“I saw you liked this LinkedIn post.”

“I noticed you commented on this thread.”

Spammers think this is personalization.

It is the opposite. It is creepy, it signals desperation, and it kills reply rates before the email is even read.

Here’s how I play it:

Email 1 Subject: {{Your company}} x {{Prospect’s company}}
Hey {{firstname}},
I am reaching out from Leed. If you ever need to grow your LinkedIn, gain credibility, build your brand, and get inbound business in the long run, feel free to reach out. I am happy to help when the timing is right.
Best, Yonathan

Email 2 Subject: {{Your company}} x {{Prospect’s company}}
Hey {{firstname}},
Hope you are doing well. Tell me if this is something you want to explore later in the year.
Best, Yonathan

Email 3 Subject: {{Your company}} x {{Prospect’s company}}
Hey {{firstname}},
I am sending you a PDF of your last LinkedIn post rewritten by us so you can see how we can help. If this is not a priority right now, I can circle back in a few months.
Cheers, Yonathan

Simple. Respectful. No fake urgency. This sequence alone has booked more meetings than most automated sequences I have seen.

To be transparent: I use this at the start of a campaign, as a testing ground. It opens conversations. The replies are genuine.

These days, all my leads come through LinkedIn. I have no need to do outreach anymore. But I still run this sequence regularly when I want to spark an interesting discussion or test a new angle.

It works. And it costs nothing but a few minutes to set up.

The 360° LinkedIn Profile & Content Audit Prompt

Yonathan Levy — Tue, 14 Apr 2026 10:44:25 GMT

Your LinkedIn profile is doing a job right now. But ask yourself, is it the right one?

Is it telling people what you do?
Is your content pointing back to a profile that makes them want to stay?
Or are both just sitting there, doing their own thing and hoping the other one picks up the slack?

You might think your LinkedIn profile and your content are two separate problems.

I can assure you, it’s not.

A great headline is like a great restaurant with a terrible menu. You got people through the door. Now what?
Same goes for great content; it means nothing if the profile it points to does not convert.

Both have to work together, at the same time.

This prompt treats both as one system.

It runs inside Claude in Chrome, which means it reads your profile live - directly off the page. Your photo, your headline, your about section, your last ten posts. You do not copy-paste a single thing. (If you have ever spent 20 minutes trying to paste your own about section into a chatbot, you know exactly why that matters.)

Before it makes a recommendation, it asks you seven questions. Your goal, your audience, the one action you want someone to take when they land on your profile. That is what makes every output specific to you. (without this part, the audit won’t make sense at all.)

What it Covers

Profile Side:

Three headline rewrites - one for search, one for conversions, one for the feed
Your about section rewritten from scratch using a Hook-Story-Offer structure
A verdict on your featured section and exactly what to pin instead
Your experience entries turned from duty lists into achievement statements
A skills strategy - what to keep, what to drop, how to get more endorsements on what matters

Content side

Every recent post is broken down individually - overperformer, average, or underperformer, with a specific diagnosis for each
An audience signal map showing what your people reading your posts respond to and what they ignore
Five post ideas built directly from what already worked
A 30-day strategic direction with three structural changes to how you write and format posts

At the end, you get a numbered action list. Every fix, in the order you should do it. Profile first - highest leverage, one time. Content second, ongoing.


You are a 360-degree LinkedIn Strategist — part profile architect, part content analyst, part audience psychologist. You are operating inside Claude in Chrome, which means you can read the user's LinkedIn profile and recent posts directly from the page — no copy-pasting required. Your job is to run a complete audit covering every layer of their LinkedIn presence: visual identity, profile copy, positioning, and content performance.



LinkedIn is the highest-leverage professional platform in the world. Over 95% of recruiters use it as a primary screening tool. Decision-makers spend 7–15 seconds on a profile before moving on. A true 360 audit treats profile and content as one system — because a great headline means nothing if the content beneath it loses the audience, and great content means nothing if the profile it points to doesn't convert.

You are reading this profile live. You have access to:
– Profile photo
– Headline
– About / summary section
– Featured section (pinned content)
– Experience descriptions
– Skills section
– Recent posts and their visible engagement (likes, comments, shares, reposts)




PHASE 1 — READ FIRST, ASK SECOND

Before asking anything, silently read the entire profile and all visible recent posts. Absorb everything you can see on the page. Then — before running any analysis — ask the user the following questions. These questions fill in context the page cannot tell you. Do not skip this phase. The audit quality depends on it.

Ask all questions in one message, numbered, conversational tone. Do not run the audit until the user has answered.

Questions to ask:

1. What is your primary goal right now?
   (job search / attracting clients / building thought leadership / growing your network / all of the above)

2. Who is your target audience?
   (recruiters in [industry] / potential clients in [niche] / industry peers / a specific role type — be specific)

3. When someone lands on your profile, what is the ONE action you want them to take?
   (send a connection request / book a call / follow you / DM you / visit a link)

4. How would you describe your current positioning in one sentence — how do you want to be known?

5. Which of your recent posts are you most proud of, and which do you think underperformed? Just name them briefly.

6. Is there anything about your profile or content you already know needs fixing but haven't gotten around to?

7. Any hard constraints? (things you cannot change, topics you won't post about, tone you must maintain)

---

PHASE 2 — RUN THE FULL 360 AUDIT

Once the user has answered all questions, run the complete audit using both what you read from the page AND the answers they gave. Never contradict what you directly observed. Never guess at metrics you cannot see — if engagement numbers are not visible, note that and work with what is there.

PROFILE AUDIT

1.1 Profile Photo
Evaluate: face framing (should fill most of the frame), background cleanliness, lighting quality, whether it matches the niche and the audience's expectations.
If it needs improving: write a Nano Banana image generation prompt the user can use to create a better version.

1.2 Headline
Diagnose the current headline specifically — what it communicates, what it misses, who it speaks to (or fails to).
Write three alternatives:
– Option A: keyword-optimised for search and recruiter scanning
– Option B: value-proposition-led, speaks to the target audience's outcome
– Option C: bold / pattern-interrupt, built for scroll-stopping on the feed
For each, explain what it does differently and when to use it.

1.3 About / Summary Section
Diagnose the current about section: hook strength, value proposition clarity, social proof, call to action.
Rewrite it using the Hook–Story–Offer framework:
– Hook: opens with a specific claim, contrarian take, or direct address to the reader's situation
– Story: establishes credibility, context, and what makes this person different
– Offer: tells the reader exactly what to do next and why it benefits them

1.4 Featured Section
State what is currently pinned.
Verdict: is it serving the goal the user stated in Phase 1?
If not: say exactly what to replace it with and why.

1.5 Experience Descriptions
Pick the most relevant experience entry visible on the page.
Before: show the current text.
Problem: explain what it is doing wrong (duty list, no metrics, no outcomes, too vague).
After: rewrite it achievement-focused with metrics where inferable or clearly placeholder-marked.

1.6 Skills Strategy
Evaluate the skills visible on the page against the user's stated goal and audience.
Output: keep / add / drop list with reasons.
Add one specific tactic for getting more endorsements on the skills that matter most.

---

CONTENT AUDIT

2.1 Post-by-Post Breakdown
For every recent post visible on the page, write one paragraph.
Lead with: Overperformer / Average / Underperformer (based on visible engagement relative to the account's own baseline).
Then give the specific diagnosis — not "good hook" but what the hook did and why it worked or failed.

2.2 Audience Signal Map
Two columns:
– What they respond to: specific topics, formats, angles, structures backed by what you observed
– What they ignore: patterns across the weak posts

2.3 Pattern Findings
Bullet points only. Concrete observations. No speculation. If the data does not support a claim, do not make it.

2.4 Strategic Direction — Next 30 Days
A. One clear strategic shift: what to do more of, less of, or stop entirely — a real change, not "keep doing what's working"
B. Three structural changes to how posts are written or formatted, based directly on observed patterns
C. Five post ideas built from overperformers — not new ideas, extensions of what already worked. For each: angle, format, and why it is the logical next step from the data.

---

IMPLEMENTATION SEQUENCE
Number every recommended action in the order the user should do them — profile fixes first (highest leverage, one-time), then content changes (ongoing).

METRICS TO TRACK
List the specific numbers to monitor over the next 30 days to measure whether the changes are working.




– You are reading the page directly. Never ask the user to paste content you can already see.
– No generic advice. Every recommendation must be tied to this specific profile, this niche, this audience, this goal.
– No softening failures. If a post failed because the hook was vague, say that.
– No hallucinating metrics. If engagement numbers are not visible, flag it and work with what is there.
– No em dashes anywhere in the output.
– No AI-sounding filler ("it's worth noting," "let's dive in," "this could potentially").
– New line after every full stop in all rewritten copy and post ideas.
– Do not run the audit until Phase 1 questions are answered.



The question phase exists because the page can tell you what is there but not why — the user's goal, intended audience, and desired action determine whether any given profile element is working or broken. A minimalist headline is a problem for a job seeker and a feature for a thought leader. Without Phase 1, the audit produces technically correct but strategically wrong recommendations. Phase 1 takes 60 seconds and makes the entire audit 10x more accurate.

The audit itself treats profile and content as one system. Profile fixes convert attention. Content fixes generate attention. Both must be aligned to the same goal and audience or they pull in opposite directions.



The user has opened their LinkedIn profile in Chrome with Claude in Chrome active. Begin by reading everything visible on the page. Then run Phase 1 — ask all seven questions in one message before doing anything else. Wait for their answers before proceeding to the audit.

Turn YouTube Videos Into LinkedIn Carousels

Yonathan Levy — Fri, 20 Feb 2026 07:32:40 GMT

Many of us stare at a blank screen trying to come up with LinkedIn content. I stopped doing that a while ago.

Instead, I let YouTube do the hard part, finding ideas that already work and I use AI to transform them into carousels. Here’s the exact process, prompts included.

The 5-Step Process

Step 1 - Find the right video

Don’t pick a random video. Go for something with high retention. Long watch times, lots of comments, 1M+ views. Copy the link. That’s your raw material.

Step 2 - Extract the key ideas

First, grab the video’s transcript. On YouTube, click the three dots under any video → “Show transcript.” Copy the whole thing.

Then paste it into your AI tool with this prompt:

Here is the transcript of a YouTube video.
I'm turning it into a LinkedIn carousel.
Read it and extract:

- The main ideas worth teaching
- Any strong opinions or clear takeaways
- Points that can stand alone as individual slides

Output the ideas as short, clear bullet points.
Do not write slides yet.
Do not add filler or explanations.

[PASTE TRANSCRIPT HERE]

You’ll get a clean list of the core ideas. Don’t skip this step because it forces the AI to think before it designs.

Step 3 - Turn ideas into a slide outline

Now take those bullet points and run this prompt:

Using the ideas extracted earlier, turn them into a clear slide outline.

Requirements:
- 6–8 slides total
- Each slide should focus on one idea only
- Write a short headline for each slide
- Add one supporting line per slide
- Order the slides so they flow logically from start to finish
- Do not design visuals.
- Stop after the outline.

Review the outline before moving on. Cut anything weak, fix the order if it feels off, and make sure the story flows from slide 1 to the last.

Step 4 - Generate the visuals

Once the outline is tight, hand it back to the Gemini with your design rules:

Using the slide outline above, generate the complete carousel.

For each slide:
- Generate the final slide content exactly as it should appear
- Include the headline and supporting text
- Keep one clear idea per slide

Design rules:
- Vertical carousel format (4:5)
- Clear visual hierarchy (headline first, support below)
- Short, scannable text
- Consistent layout across all slides
- No clutter, no dense paragraphs

Style rules:
- Brand colors: [your hex codes]
- Simple shapes, dividers, or visual elements where helpful
- Keep the design clean and readable

Swap in your own brand colors, and you’ve got a near-complete carousel.

Step 5 - Final polish in Canva

AI gets you about 80% there. The last 20% is just taste. Open the output in Canva, swap icons and shapes to match your style, apply your branding, and you’re done.

Pro tip: If any AI-generated images have watermarks, remove them with Canva’s Magic Eraser, a simple crop, or any basic photo editor.

That’s the whole system. A 20-minute YouTube video becomes a carousel in under an hour, and you didn’t have to come up with ideas from scratch.

Claude Co-Work Workflows for LinkedIn

Yonathan Levy — Tue, 03 Feb 2026 08:33:59 GMT

I promised workflows, so here they are.

These are the exact prompts I use with Claude Projects to write content faster. Nothing fancy. Just copy-paste-done.

Each workflow solves one problem: making old posts searchable, breaking down why certain posts work, turning one idea into multiple posts, etc.

Let’s get into it.

Workflow 1: Make your posts searchable

Please create a searchable knowledge base from all my LinkedIn posts:

1. Use Claude in Chrome to access my LinkedIn profile at [YOUR LINKEDIN URL]
2. Read through all my posts (go back as far as you can)
3. Extract every post and organise them into a structured document with:
    - Posts grouped by theme/topic
    - Each post tagged with relevant keywords
    - A table of contents for easy navigation
    - Your analysis of recurring frameworks or ideas I use
    - Topics I've covered frequently vs. gaps I haven't addressed
4. Format it as a searchable markdown document
5. Include post dates so I can see how my thinking has evolved

The goal is to make it easy for me to search "what did I say about X" and find it immediately, plus identify patterns in my own thinking.

Workflow 2: Turn viral posts into repeatable formats

Please analyze high-performing LinkedIn posts and help me write in that structure:

1. Use Claude in Chrome to read these LinkedIn posts:
[PASTE 3-5 POST URLS]
2. Break down each post's structure:
    - How does the hook work? (first 1-2 lines)
    - How do they build the argument or story?
    - Where do they pivot or add tension?
    - How do they land the payoff or CTA?
3. Identify the common patterns across these posts that make them effective
4. Now take my idea: [YOUR TOPIC/IDEA]
5. Write it using the same structural principles you identified, but in my voice and for my audience: [DESCRIBE YOUR AUDIENCE]
6. Give me 2-3 versions using different structures from the posts you analyzed, and explain why each structure works for this specific idea.

Workflow 3: Turn one idea into five posts

I have one core insight I want to turn into multiple LinkedIn posts without being repetitive.

My core insight: [YOUR INSIGHT - can be rough notes, a paragraph, whatever you've got]

My audience: [WHO YOU'RE WRITING FOR and what they care about]

Please create 5 completely different posts that explore this insight from different angles:

1. The contrarian take - challenge conventional wisdom related to this insight
2. The tactical how-to - make it actionable with specific steps
3. The personal story version - use narrative/experience to illustrate the point
4. The data/example-driven version - use evidence or case studies
5. The "here's what most people miss" version - surface the non-obvious implication

For each post:

- Make sure it stands alone (no "Part 2" references)
- Keep my voice
- Include a strong hook
- Make the structure clear

Also suggest which order I should post these in and why.

Workflow 4: Turn conversations into content

Please build me a LinkedIn content calendar based on what I'm already saying in meetings and conversations:

I'm uploading: [meeting transcripts / Slack threads / email exchanges / voice memos where I explained ideas well]

Please:

1. Read through everything and identify the strongest, most interesting ideas I'm already articulating
2. Pull out 10-15 distinct post concepts organized by theme
3. For my top 5 ideas, write full first drafts
4. Create a content calendar showing:
    - When to post what (spread across 2-4 weeks)
    - Why this sequence makes sense
    - Which ideas build on each other vs. stand alone

Focus on ideas I've already validated by explaining them to real people. Don't just summarize what I said - turn it into posts optimized for LinkedIn.

My audience: [YOUR AUDIENCE]
My goals: [What you want to be known for / what conversations you want to start]

Workflow 5: Mine long content for posts

Please extract every usable insight from this podcast and turn it into LinkedIn content:

Podcast: [YOUTUBE LINK / UPLOAD TRANSCRIPT]

What I want you to do:

1. Read/watch the entire thing
2. Extract every meaningful insight, organized by theme
3. Identify 8-10 ideas that would work as standalone LinkedIn posts
4. For each idea, give me:
    - The core insight (1-2 sentences)
    - Why it matters to my audience
    - A first draft of the post
    - Timestamp reference (so I can verify or add more context)

Focus on:

- [Contrarian takes / surprising moments / tactical advice / personal stories / frameworks mentioned / whatever matters most to you]

My audience: [YOUR AUDIENCE]

Don't just transcribe - turn these into posts that work on LinkedIn. Extract the best thinking and make it shareable.

LinkedIn profile Audit System (PROMPT)

Yonathan Levy — Sat, 24 Jan 2026 19:44:41 GMT

Why most LinkedIn profiles fail in under 10 seconds

Recruiters don’t “read” LinkedIn profiles.
They scan them.

In practice, you have between 7 and 15 seconds to answer one question:

“Is this person worth my attention?”

Most profiles fail because:

the headline is vague,
the summary talks about tasks instead of value,
the experience section reads like a job description,
and there is no content strategy behind the profile.

To fix that, you don’t need tricks.
You need structure, positioning, and recruiter psychology.

So I built a single prompt that forces AI to behave like an elite LinkedIn profile strategist instead of a generic copy generator.

You can paste it into ChatGPT and get:

a full profile audit,
concrete rewrites,
and a content plan aligned with your career goals.

Here is the exact prompt:

You are an elite LinkedIn Profile Strategist with expertise in personal branding, talent acquisition, and digital professional presence. Your specialization is transforming underperforming LinkedIn profiles into powerful career advancement tools.
LinkedIn has become the premier platform for professional opportunities, with over 95% of recruiters using it as a primary screening tool. The average decision-maker spends only 7-15 seconds scanning a profile before deciding to engage or move on. Despite this, most professionals have profiles that fail to capture attention or communicate their true value proposition. The difference between a mediocre and outstanding LinkedIn profile can significantly impact career trajectory, salary negotiations, and access to premium opportunities.
Conduct a comprehensive audit of the user’s LinkedIn profile, analyzing all key elements:
1. First, request the user’s current LinkedIn information including:
- Current headline
- About section/summary
- Experience descriptions
- Skills section
- Recent activity/content shared
- Current goals (job searching, networking, thought leadership, etc.)
- Target audience (recruiters, clients, industry peers)
2. Evaluate each profile element against industry best practices, identifying:
- Headline effectiveness and keyword optimization
- Summary impact and value proposition clarity
- Experience descriptions (achievement focus vs. duty lists)
- Skills relevance and endorsement strategy
- Content strategy gaps
- Visual elements and profile completeness
3. Provide actionable recommendations for improvement:
- Create 3 powerful headline alternatives with explanation
- Rewrite their summary using the “Hook-Story-Offer” framework
- Transform one experience description from task-focused to achievement-focused
- Suggest optimal skills arrangement and endorsement strategy
- Develop a 30-day content calendar with 5 specific post ideas tailored to their industry
4. Explain the strategic rationale behind each recommendation, citing LinkedIn algorithm preferences and recruiter psychology.
- Avoid generic advice; all recommendations must be specifically tailored to the user’s industry, career level, and goals
- Focus on authentic positioning rather than keyword stuffing or inauthentic tactics
- Do not request sensitive personal information beyond what would typically appear on a LinkedIn profile
- Ensure all recommended content ideas align with the user’s stated professional brand
- Do not make unrealistic promises about guaranteed job offers or specific salary increases
Provide your analysis in this structured format:
LINKEDIN PROFILE AUDIT
Current Profile Strengths:
[List 3-5 positive elements of their existing profile]
Critical Improvement Areas:
[Identify 3-5 specific weaknesses holding back their profile performance]
Strategic Recommendations:
1. Headline Transformation:
[3 alternative headlines with explanation]
2. Compelling Summary Rewrite:
[Transformed summary using Hook-Story-Offer framework]
3. Experience Description Optimization:
[Sample before/after transformation of one experience entry]
4. Skills & Endorsements Strategy:
[Specific recommendations for skills section]
5. Content Strategy Blueprint:
[5 specific post ideas with optimal posting cadence]
Implementation Priority Guide:
[Numbered list of actions in recommended sequence]
Performance Measurement:
[Specific metrics to track profile improvement]
The audit approach uses a systematic analysis of all LinkedIn profile elements against established best practices from talent acquisition research. The recommendations leverage psychological principles of attention capture, value proposition communication, and social proof to maximize profile effectiveness. The structured output ensures actionable implementation rather than overwhelming the user with general advice.
Start by asking the user to enter the details as described on the seciton, item 1. Then wait for the user to provide their specific LinkedIn profile information.

Grok + LinkedIn

Yonathan Levy — Tue, 06 Jan 2026 11:58:14 GMT

Prompt 1: Find ideas people are already arguing about

Find recent X posts about [topic/niche] that each have at least 5–10 replies.
Ignore viral posts and ignore posts with zero engagement.
From these posts:
– Identify 3 emerging opinions people are actively responding to
– Explain why each opinion is getting replies
– Highlight what part of the idea feels unresolved or emotionally charged
– Suggest how this could be turned into a LinkedIn post that sounds early, not recycled”

Prompt 2: Knowing the objectives before you post

I’m planning a LinkedIn post with this opinion: [your take].
Find:
– The strongest arguments against it
– What skeptics would say in the comments
– Where this take might sound naive or oversimplified
Then suggest how to sharpen the idea without softening it.

Prompt 3: Validate whether an idea has signal or is already exhausted

Find recent X posts related to [your topic] where each post has at least 5 replies.
Cluster these posts by idea, not keywords.
For each cluster:
– Estimate how frequently this idea is already being repeated
– Identify which angles feel overused vs which feel underexplored
– Explain what people are still reacting to and what they’re ignoring
– Suggest one LinkedIn angle that avoids repetition while staying relevant

Prompt 4: See how problems show up in conversations

Find X posts discussing [problem or tool] with at least 5 replies each.
From both the original posts and replies:
– Extract exact phrases that sound frustrated, tired, or blunt
– Do not rewrite or clean the language
– Group phrases into themes (pressure, confusion, fatigue, scepticism)
– Suggest how one phrase could be used directly as a LinkedIn hook

Prompt 5: Say what people think but won’t post

Scan X discussions on [topic] with posts that have a minimum 5–10 replies each.
Identify 3 perspectives that:
– Show up in replies but rarely as original posts
– Are expressed indirectly (sarcasm, frustration, side comments)
– Feel socially risky or ‘uncool’ to say on LinkedIn
For each perspective:
– Explain why people hesitate to post it publicly
– Identify the underlying truth being hinted at
– Suggest a LinkedIn post angle that states it clearly but thoughtfully

Follow me on Substack for more!

Subscribe now

SDRs: Never Start a Month at Zero

Yonathan Levy — Thu, 27 Nov 2025 08:29:46 GMT

SDRs wait for the new month to begin before they start booking meetings. That habit kills the possibility to overachieve concistently.
Here is the system that lets you enter every month with meetings already scheduled, less pressure, and a compounding pipeline.

THE MISTAKE MOST SDRS MAKE

SDRs use to start prospecting once the new month starts.
That is already too late.

Your goal is not to force immediate meetings.
Your goal is to create a steady stream of future conversations.

The rule is simple.
Success means having meetings already scheduled before Day 1.

Nothing reduces pressure more than that.

WHY SIX MONTHS AHEAD WORKS

When a prospect says the timing is bad, most SDRs drop the lead or toss it into a forgotten CRM list.

The better approach is simple.
If they say five months, book a slot in six.

From my own experience, the meetings booked in advance turn out to happen and be qualified at 65%.

Prospects agree to future meetings much more easily because the commitment costs them nothing right now.
And for you, it creates future momentum.

Even if only part of these meetings are held, your pipeline compounds.
You stop living in monthly panic mode.
You start building months ahead.

THE EMAIL THAT MAKES IT POSSIBLE

You do not need a clever template.
You need a clean human email that gets responses.

Why it works:

Easy to open
No pressure
Actually helpful
Builds long-term pipeline, not just this month

Simple messages from real humans perform better than anything overly polished.

THE REAL BLOCKER IS NOT THE WRITING

The real blocker is volume.
To book meetings ahead, you need enough daily outbound to plant future conversations without killing your domain.
That is why Maildoso matters.

Ten mailboxes, fifty emails each, about five hundred safe and human emails per day.
You keep your best human message.
Maildoso gives you the scale.

HOW PIPELINE STARTS TO COMPOUND

Once you send simple human emails every day at consistent volume, outbound becomes predictable.

Some prospects reply right away.
Some reply in a few weeks.
Some reply in a few months.
Some show up to the meeting you booked six months earlier.

Every email plants a seed.
They do not sprout at the same time.
But they all grow.

This shift changes your entire relationship with quota.
You stop sprinting every month.
You start compounding.

Test Mailodoso now

HERE IS THE COMPLETE 3 EMAIL SEQUENCE

10 days between each email

Short, human, zero pressure. Perfect for booking future meetings.

Email 1

Subject: Subject: {{Your company}} x {{Prospect’s company}}
Hey {{firstname}},
I am reaching out from Leed.
If you ever need to grow your LinkedIn and gain credibility, build your brand, and get inbound business in the long run, feel free to reach out.
I am happy to help when the timing is right.
Best,
Yonathan

Email 2

Subject: Subject: {{Your company}} x {{Prospect’s company}}
Hey {{firstname}},
Hope you are doing well!
Tell me if this is something you want to explore later in the year.
Best,
Yonathan

Email 3

Subject: {{Your company}} x {{Prospect’s company}}
Hey {{firstname}},
I’m sending you a PDF of your last LinkedIn post rewritten by us so you can see how we can help.
If this is not a priority right now, I can circle back in a few months.
Cheers,
Yonathan

Test Mailodoso now

The New Cold Email Game: How to Win With Quality at Scale

Yonathan Levy — Thu, 13 Nov 2025 14:15:39 GMT

Cold email is not dead. Bad email is. Here is the 2025 playbook for sending 300 targeted emails a day with high deliverability and real personalization.

Caption: The workflow behind quality at scale.
Alt text: Diagram showing a cold email process with CRM data, Perplexity research, and Maildoso mailboxes.

There is a loud debate in sales right now. Some say cold email is finished. Others say it is still the best channel in the world. The truth is simple. Cold email is not dead. Bad cold email is.

What worked in 2023 does not work today. Sending 500 messages from one inbox and hoping something lands is gone. Deliverability collapses. Domain reputation sinks. And once it drops, the entire team pays the price.

Yet volume matters. Outbound works only when your message reaches enough of the right people. So the real question is not volume or relevance. The real question is how to combine both without destroying deliverability.

The modern cold email workflow solves this. You get relevance powered by AI research, combined with safe scaled sending.

This article breaks down the full system.

Article sponsored by Maildoso

Article sponsored by Maildoso

Test Mailodoso now: getmaildoso.com/?aff=YonathanCheck

The shift: relevance is the new multiplier

For years, cold email was a numbers game.
Spray. Pray. Book meetings.

Everyone used the same templates, the same scrapers, the same tricks. Buyers were flooded with generic emails. Filters became stricter. Inbox providers rewarded authentic, human communication.

Today the winning equation is different.

Fewer emails.
Stronger targeting.
Higher relevance.
Better replies.

Every successful outbound motion starts with context. You need to speak to a real scenario the buyer is experiencing, not a random guess.

Relevance alone is not enough. Without volume, you are running a content project, not an outbound engine.

The challenge is clear. Scale personalized outbound without sounding like a template and without burning your domain.

Step 1: Start from your CRM, not from a scraper

Most reps start with scraped lists from random databases. It feels fast, but it produces the weakest accounts. You end up emailing companies that will never buy.

High quality outbound starts in the CRM.

Pull a list of your real ICP accounts.
Export them into a CSV.
Focus on accuracy, not size.

This is your raw material. Everything else builds on it.

Step 2: Feed the list to Perplexity for real research

Upload your CSV into Perplexity. You get company intelligence at scale in minutes.

Perplexity scans the web for every account.
It extracts funding rounds, hires, product launches, expansions, partnerships, and market moves.
It ranks companies by momentum and opportunity.

Your static list becomes dynamic insight.

Every account gains a reason to contact them.
Every email starts with context.
Every message feels researched and intentional.

Most teams skip this step. They rely on gut feeling. But the gap between generic personalization and true context is massive. Buyers feel it instantly.

Step 3: Turn data into relevance

Research does not write your email. You need to turn insight into a strong opening line.

The formula is simple.

Pick one concrete fact.
Link it to a problem or opportunity you solve.
Frame your message around that connection.

No long intros.
No robotic personalization.
No forced LinkedIn references.

A single relevant sentence sets the tone and proves you did real work. Reply rates rise sharply.

But better emails take time. Manual research does not scale. And domain limits prevent you from sending volume.

This is where most teams get stuck.

The volume problem: deliverability is the new ceiling

Email providers now cap cold email sending. Crossing 50 cold emails a day per mailbox creates high risk. If you push past this limit, reputation damage follows.

Most teams settle for low volume. That kills pipeline.

The issue is not volume itself. The issue is how to maintain volume without harming deliverability.

The answer is mailbox distribution.

Step 4: Use Maildoso to scale safely

Maildoso brings back safe high volume.

You connect multiple warmed mailboxes.
You distribute sends across them.
You protect your domain while increasing output.

Six mailboxes.
Fifty emails each.
Three hundred targeted emails per day.

Every email uses real context from Perplexity.
Every mailbox stays within safe limits.
Every domain stays clean.

Personal.
Scalable.
Predictable.

You get the volume of 2023 and the personalization of 2025.

Test Mailodoso now!

BONUS:

The formula

That makes outbound work again

(Using Perplexity)

Perplexity gives you quality.
Maildoso gives you volume.

Together, they create cold emails that feel human, relevant, and researched without manual burden.

This gives reps a new edge.
This gives founders leverage.
This gives teams a scalable outbound engine that respects modern rules.

Instead of sending templates at scale, you send context at scale. That is the shift.

PROMPT:

You are a sales research assistant.

Task:
You receive a CSV of companies as input.
For each company, you must search the web and find the freshest, high-impact signal.

Valid signals:
- Funding: rounds, investors, IPO, major grants
- Hiring: big hiring push, key exec hires, new teams
- Product: launches, major feature releases, pivots
- Partnerships: strategic partnerships, integrations, resellers
- Expansion: new markets, new countries, new verticals
- Offices: new HQ, new office openings or closures

Rules:
- Use only verified public information.
- Prefer news from the last 90 days.
- Ignore low impact noise such as generic blog posts, small content updates, minor marketing news.
- If you find several signals, pick the single most relevant and recent one.
- If you find nothing meaningful, write “No strong recent signal found” and leave the link field empty.

Output:
Return a ranked table from highest to lowest relevance.

Columns:
1) Company name
2) Key signal found
3) Cold email intro sentence
4) Source link

Cold email intro sentence:
- 1 sentence.
- Must feel specific and recent.
- Mention the exact event (funding round, launch, partnership, new office, etc.).
- Sound natural, like the first line of a cold outbound email, not like a headline.

Ranking:
- Rank companies where you found big, very recent events at the top.
- Then companies with strong but slightly older events.
- Then companies with weaker or older signals.
- Last: companies with “No strong recent signal found”.

Now read the CSV and produce the ranked table.

Use research for a better result.

Test Mailodoso now: getmaildoso.com/?aff=YonathanCheck

Why this matters for 2025 and beyond

The outbound world will keep tightening. Providers will add rules. Buyers will expect more relevance. AI will keep raising standards.

Teams that rely on brute volume will disappear.
Teams that rely on slow manual personalization will stay small.
Teams that combine both will win.

Outbound is no longer about hacks.
It is about systems that compound.
Systems that mix precision with reach.
Systems that scale without burning domains.

Get this right and your outbound stops feeling like noise. It feels like service.

If you found this useful, subscribe for weekly breakdowns on outbound, personalization, and high leverage sales workflows.

The hidden layer of email success

Yonathan Levy — Thu, 30 Oct 2025 12:02:35 GMT

When people say “email is dead”, they mean lazy email is dead.

The kind that gets sent by the thousand, from a cold domain, to a cold list, with no care.

But high-performing email is alive.
And it’s winning quietly.

The real game isn’t about the copy.
It’s about the infrastructure that supports it.

The domains.
The warm-up system.
The hidden mechanics under every “sent” button.

That’s where 90% of results come from.
And that’s what most sales teams still ignore.

What changed

Email in 2025 isn’t what it was even a year ago.
Filters are smarter.
ESP scores are public.
And deliverability is no longer just a “technical thing”, it’s a growth moat.

Your cold domain reputation is now your credibility score.
Mess it up once, and every next campaign costs more.

This is why tools like Maildoso matter.
They treat your sender setup like a real asset, not an afterthought.

What I learned after 1 month

After switching my infrastructure, something interesting happened:
I started thinking about email like SEO.

→ Every domain is an ecosystem.
→ Every interaction trains the filters.
→ Every bounce, click, and reply affects your future visibility.

The more consistent and clean your setup is, the easier it gets to land in the inbox — even before you write a word.

I stopped obsessing over subject lines.
I started obsessing over sender health.

The next frontier

This shift is bigger than deliverability.
It’s redefining what “scale” means in outbound.

Before: send more, hope for replies.
Now: build trust at scale, and watch the system self-compound.

Smart infrastructure compounds like brand.
It grows in value over time.
That’s what makes this new era of cold email exciting.

If your inbox feels like a black hole, the issue isn’t your writing.
It’s your foundation.

Get that right, and every email starts working harder for you.

You don’t need luck. You need a clean system.

I’ve spent years inside top tech startups, studying what actually makes people reply.
After testing hundreds of sequences, I built one that performs the best.
It’s simple, fast, and battle-tested.
Here’s the exact sequence I use.

Here’s my email sequence to use:

Email 1 - Intro

Subject: Your company name x prospect’s company

Hey {{first_name}},
Hope you’re doing well.

I saw you do {{what_they_do}}.
We help teams like yours {{what_you_help_with}}, mainly around {{3_short_bullets_about_value}}.

Curious: how do you currently handle {{specific_pain_point}} internally?

Best,
{{your_name}}

Email 2 - Context

Subject: Your company name x prospect’s company

Hey {{first_name}},
I’ll give you a tiny bit more context.

What we do is {{one_line_explanation}}, which helps {{role_or_team}} save time and {{impact}}.
It’s simple to test, no big setup.

Does that sound relevant for {{company_name}} right now?

Best,
{{your_name}}

Email 3 - Follow-up

Subject: Your company name x prospect’s company

Hey {{first_name}},
Not sure if my last note reached you.
The reason I’m insisting is because {{pain_point}} is something we see a lot with {{similar_companies}}.

Happy to walk you through how others deal with it.
Would that be useful?

Best,
{{your_name}}

Email 4 - Breakup

Subject: Your company name x prospect’s company

Hey {{first_name}},
No worries if it’s not a fit.
Just tell me if I should stop bugging you. I don’t want to drive you crazy.

If it’s a timing thing, I can follow up later.

Take care,
{{your_name}}

Why I Write My Cold Emails This Way

Most sales emails sound fake.
Too pushy, too polished, too robotic.

You can feel when someone’s trying to sell you something.
And that’s exactly why most people don’t reply.

So I stopped writing like a salesperson.
I started writing like a human.

That’s why my emails begin with simple lines like “Hope you’re doing well” or “I saw you do X.”
Because that’s how real people talk.

Then I add curiosity: “How do you currently handle X internally?”
It’s not a trick. It’s respect.
I’m showing I actually care about their context before pitching anything.

I also end my emails with honesty: “Tell me if it’s not a fit, I don’t want to drive you crazy.”
That sentence alone gets me more replies than any call-to-action ever did.

This make my email sounds human and a no is yes for my CRM to be clean and filled of data.

The point isn’t to sell in the first email.
It’s to start a real conversation.

People buy from people who sound real and relevant .
That’s the only “strategy” that still works.

Last thing from experience, do not try fancy subjects. Your company x prospect’s company is enough.

From Research to LinkedIn Posts in 15 Minutes

Yonathan Levy — Fri, 24 Oct 2025 09:50:27 GMT

If you want ideas that are recent, specific, and ready to post, this combo is unbeatable.
Perplexity finds the latest data. EasyGen turns it into high-performing LinkedIn posts.

Step 1 — Open Perplexity

Go to Perplexity.ai.
Open a new thread in Web mode.
Get your topic ready. Keep it short and clear (you’ll paste it on one line).

Step 2 — Paste the prompt and fill your topic

Copy this prompt into Perplexity, then replace the line starting with TOPICS: with your subjects (all on one line).

Act like a senior B2B sales and AI analyst who is also a LinkedIn content strategist for GTM teams (SDRs, AEs, RevOps). You write in a clear, impactful, educational voice, no jargon, no fluff, no emojis, no hashtags.

TOPICS: Sales.ai, AI for Sales, SDR↔AE collaboration   # ← Replace this with your own topics (keep one line)

Objective:
- Research and synthesize the latest trends, news, and insights on the above topics.
- Create LinkedIn-ready content ideas that are short, specific, and outcome-oriented.

Ground Rules:
- Web-browse and prioritize sources from the last 6–12 months.
- Verify facts and cite credible sources with inline markdown links [Source].
- Avoid speculation. Label predictions clearly.
- Use plain English and concrete examples.

Deliverable:
Produce exactly 10 complete ideas.  
Each idea must follow this exact order and structure:

1) Title: (≤12 words, clear benefit, use “Here’s how…”, “Here’s why…”, “In 2026…”)  
2) Hook: (1–2 sentences, short, human, no jargon)  
3) Re-hook: (1 sharper, more contrarian version of the Hook)  
4) Paragraph: (2–4 lines; summarize the insight, show payoff, include 1–2 specifics like a stat, tactic, or framework, invite curiosity)

Formatting Rules:
- Output in clean markdown.
- Do not separate parts by type. Each idea must include Title, Hook, Re-hook, and Paragraph.
- No commentary or explanations outside the 10 ideas.

Quality Checklist:
- Every idea must link directly to the chosen topics.
- Replace vague claims with data or examples.
- Keep it short and consistent.
- No duplication or unverifiable statements.

Template to repeat exactly 10 times:

1) Title:  
   Hook:  
   Re-hook:  
   Paragraph:

Take a deep breath and work on this step-by-step.

Pro tip:
If Perplexity outputs sections in the wrong order, type this line at the top before running:
Print exactly 10 ideas following the structure Title, Hook, Re-hook, Paragraph.

Step 3 — Review your results

Keep 3 to 5 ideas that fit your audience.
Add your real-world experience or data.
Copy them into EasyGen for the next step.

Step 4 — Generate your posts with EasyGen

Go to EasyGen.
It’s a powerful AI platform that helps you turn structured ideas into ready-to-publish LinkedIn posts.

Why I recommend EasyGen

EasyGen is the tool that helped me start posting consistently on LinkedIn.
It turns your ideas into clear, impactful posts that actually sound human.
You can create hooks, structure carousels, or expand short notes into long-form posts.

🎁 Get a discount for EasyGen Coupon code: YONATHAN
Use this link and code to get a discount and full access.

Step 5 — Final polish and post

Check that your post delivers on its promise.
Add one specific example or data point.
End with a simple CTA (comment, DM, or takeaway).
Publish and reply to early comments within 30 minutes.

Example topic line for Perplexity

When you start a new topic, replace the TOPICS line like this:

TOPICS: AI in Recruitment, Talent Intelligence, Employer Branding

Subscribe now

Thanks for reading! This post is public so feel free to share it.

The Hardest Lesson in Sales: Protect Your Composure

Yonathan Levy — Tue, 21 Oct 2025 07:38:25 GMT

One rude prospect does not get to ruin your day. If you want to last in sales, you must master something deeper than your pitch: composure. Here is how to build it.

Most salespeople learn this the hard way.

You spend hours writing tighter openers. You tweak tone and timing. Then one rude prospect cuts you off and the whole block derails.

That is the real problem.

It is not your script that breaks you. It is the inability to reset fast.

The real game in sales

If you want to last, you need grit you can repeat.
You will get hung up on. You will hear no. You will be told never call again.
Five minutes later, you must dial with a clean mind.

That is what separates good from great.
Pros do not carry the last call into the next one.
They breathe. They reset. They treat every dial like day one.

What wins in 2026

Automation and AI are table stakes.
What wins is emotional endurance.

AI can handle grunt work. Tools can score leads and log notes.
Composure is still your job.

Composure turns every objection into data.
Composure turns the next dial into a real shot.
Composure turns rejection into resilience.

This is not a buzzword. It is a skill you can train.

The blueprint for emotional endurance

1) Breathe and reset after every call
Do not drag the last no into your next hello. Stand up. One slow inhale for four counts, hold for two, exhale for six. Repeat three times. Sit and dial.

2) Treat every call as a new world
Forget the last voice and outcome. New person, new context, new chance. Let the present prospect be the only data that matters.

3) Build emotional endurance
Track your bounce back time. How many seconds from hang up to calm breath. Shrink it over time.

4) Automate what you can
Let AI handle research, enrichment, logging, reminders, and first drafts. Free your mind for conversation, judgment, and tone.

5) Turn rejection into data
Every no hides a clue. Timing, budget, authority, fit, priority, trust. Tag it, learn from it, move on.

6) Protect your composure
Guard it like gold. Lose it and you sound like everyone else. Keep it and you stand out as the calm, credible voice.

What composure looks like in the moment

Prospect cuts you off. You smile, even if they cannot see it.
You say, got it, sounds like I caught you at a bad time. Want me to try later this week or should I send a one line email?
They answer. You tag the outcome. You breathe. You move to the next call.
No drama. Just process.

Practical tools you can start today

The 90 second reset ritual
Stand up. Look at something far away. Breathe four in, two hold, six out. Repeat three times. Shake your hands. Smile. Call.

After call note formula
Four lines in your CRM:
Trigger: objection or event
Response: what you said
Learning: what to test next time
Next test: exact line you will try

Example
Trigger: Not interested, busy Q4
Response: Totally fair, is Q1 planning yours or ops?
Learning: Timing objection, open to planning
Next test: Offer 12 minute planning slot with agenda

Patterns appear within a week.

The Composure Score
Track three numbers each day:
Bounce Back Time: average seconds from hang up to calm
Fresh Start Count: times you noticed a spiral and reset
Call Quality: rate five calls for calm, curiosity, clarity on a 1 to 5

AI as your pit crew, not your driver

Give tools the heavy lifting so your mind stays light.
Pre call brief: auto generate three bullets from CRM and LinkedIn.
Call notes: record and summarize, then paste your four line note.
Follow up: 90 word recap with two time options.
Objection library: store winning lines by tag and practice daily.

Use tech to reduce load. Spend your energy on tone, timing, and listening.

Scripts for your mindset

Keep these near your monitor.
Neutral reset: that happened, now next call.
Curiosity cue: what would make this useful for them today.
Detachment: my job is not to close, my job is to find truth fast.
Self compassion: it is hard, I am learning, dial again.

Team rituals that build endurance

Two minute drill: after a rough block, everyone resets and breathes.
Tag sync: five minutes daily to share one objection tag and one line that worked.
Coach the call, not the rep: coach the moment, not the person.
Golden first call: protect the first dial of the day. No email. Call within five minutes of sitting.

Culture makes composure easier. Make it visible.

What to implement this week

Create your Rejection Ledger and Composure Score.
Write the four line note template in your CRM.
Build five objection tags: timing, budget, authority, priority, fit.
Set one 45 minute dial block with no tabs.
Place a reset card next to your keyboard.
Choose one AI workflow to automate: pre call brief or follow up draft.
Book a 15 minute weekly review to scan patterns and pick one test.

The mindset you cannot outsource

You cannot delegate composure. You cannot automate grit.
Emotional endurance is not being tough all the time. It is resetting fast and showing up fresh, again and again.

The reps who win in 2026 will not only have strong tools. They will have the calm to hear the truth, the curiosity to learn from it, and the courage to make the next call without baggage.

This is not a list to copy. It is a system to run.

If this helped, subscribe for weekly playbooks that turn pressure into process.

Share it with your team so everyone can protect composure and win the next dial.

Less Admin, More Selling: The AI Playbook for Reps

Yonathan Levy — Mon, 20 Oct 2025 12:41:36 GMT

Great salespeople are not replaced by AI. They are powered by it. Use AI to clear inbox noise, fix CRM hygiene, and prep faster. Spend the saved hours on calls, relationships, and deals.

AI does not close your deals. It does not build trust or read the room. It cannot replace your voice or your judgment. What it can do is remove the work that keeps you from selling. Think inbox triage, CRM updates, and call prep. Take that off your plate and you get hours back every week. The best reps use those hours on real conversations.

Below is a short guide you can put to work today.

1) Automate the grind

Inbox triage

Your inbox should not control your day. Set AI to sort and flag messages. Let it draft first passes for routine replies. You focus on real leads and real questions.

Practical idea:

Create labels or folders for Hot leads, Warm leads, and Everything else.
Have AI route intros, renewals, and support to the right queues.
Use AI to draft a reply, then add your voice in two sentences.

Sample prompt:
“Draft a short reply that confirms interest, asks one clarifying question, and proposes two call times. Keep it friendly and under 120 words.”

CRM hygiene

Clean data is oxygen for a pipeline. Use AI to log calls, summarize notes, and update fields after each meeting. No more end-of-day backlog. No more missed next steps.

Practical idea:

After each call, ask AI to fill in contact role, deal stage, next action, and due date.
Use an automation that creates tasks for follow-ups and assigns them to you.

Sample prompt:
“Summarize this call in three bullets. One pain, one desired outcome, one next step with a due date. Update deal stage if needed.”

Call briefs

Walk into every meeting with context. AI can pull recent company news, the prospect’s LinkedIn activity, and notes from your last call. You spend two minutes scanning, not thirty minutes researching.

Practical idea:

Standardize a one-page brief template. Company snapshot, people, recent news, last call recap, likely objections, next step.

Sample prompt:
“Create a one-page call brief for Acme Corp. Include last meeting notes, new leadership updates, and three tailored discovery questions.”

2) Support your pipeline

Research in seconds

Stop juggling tabs. Ask AI for the info you actually need. Recent funding, a new product launch, a hiring spike, or a quote from the buyer’s last podcast. You get the signals that matter and skip the noise.

Sample prompt:
“Find three timely insights about Acme Corp that could justify a conversation about cost savings in Q4.”

Drafts and proposals

Blank pages slow you down. Let AI write the first draft. You keep the voice and the strategy. You cut what is not helpful and tighten what is.

Practical idea:

Save proven email frameworks as prompts. First touch, nudge, breakup, renewal check-in, and mutual action plan.

Sample prompt:
“Draft a proposal summary in plain English. One paragraph on goals, one on the plan, one on timeline, one on price. Keep it under 250 words.”

Objection handling

Feed AI with your win and loss notes. Ask for response options that match your positioning. You get ideas in seconds and pick the one that fits this buyer.

Sample prompt:
“Give me three concise responses. We already have a vendor. Include a question that opens the door to a pilot.”

3) Keep your outreach human

AI can find facts and write drafts. You still build the relationship. Use AI to scale the research and reminders. Then add a line that only you can write.

Personalization that lands

Ask AI for one human hook. Something the buyer actually cares about. A recent post, a shared connection, or a goal from their earnings call.

Sample prompt:
“Write a 100-word opener that references this buyer’s new role and their focus on churn reduction. Keep it specific and respectful.”

Follow-ups that never slip

Let AI remind you when to nudge. Have it draft the follow-up and track replies. You stay present without dropping the ball.

Sample prompt:
“Draft a polite follow-up that adds one new insight from Acme’s latest blog post. Keep it to five sentences.”

Scheduling without the back and forth

AI can propose times, handle time zones, and send reminders. You spend your energy on the conversation, not the calendar.

The real win

AI will not sell for you. It will give you more time to sell. Top reps report more time on calls, faster follow-ups, and cleaner pipelines. The compounding effect is real. Less grind. More growth. Fewer dropped threads. More closed won.

If you feel behind, start small and build momentum.

A simple 3 week rollout

Week 1: Inbox and follow-ups

Turn on AI triage and canned replies.
Set daily blocks for deep work.
Add a two-sentence personal touch to every AI draft.

Week 2: CRM and call notes

Use AI to log calls and update stages.
Standardize a three-bullet call summary. Pain, outcome, next step.
Create a saved prompt for post call updates.

Week 3: Briefs and proposals

Adopt the one page call brief.
Use AI for proposal first drafts.
Build a small objection library and keep improving it.

Quick tool stack ideas

These are examples, not endorsements. Features change quickly. Pick what fits your stack and security needs.

Inbox and writing: Gmail with Workspace AI, Outlook with Copilot, Superhuman AI, Grammarly for tone checks.
CRM automation: Salesforce Einstein, HubSpot AI, Pipedrive Smart Docs.
Call prep and notes: Gong, Chorus, Fireflies for summaries and next steps.
Prospect research: Clay, Apollo, ZoomInfo, LinkedIn Sales Navigator.
Docs and proposals: Notion AI, Google Docs with AI, PandaDoc or DocuSign for templates.
Scheduling: Calendly, Reclaim, Motion.

What to measure

Pick three simple metrics and watch them weekly.

Time spent selling per rep, hours on calls, and meetings.
Follow-up time, first response, and next step completion.
CRM completeness, required fields filled, and next step set.

When those move in the right direction, pipeline health follows.

Try one change this week. Turn on AI triage or automate your post-call notes. Then tell me what you did and what time it saved. If you want, reply with the one task that slows you down, and I will send back a ready-to-use prompt.

Yonathan Levy

Stop asking top performers for their script

Seven meetings a month, and email was not going to get me there

I listened to every single one

The most uncomfortable moment of my career

What I eventually worked out

The part nobody puts in the LinkedIn post

Granola

Why the phone and the watch change the shape of this

The workflow

The prompts

1. The follow ups you dropped

2. What actually books

3. Where you lose control

4. The gap between what you were taught and what you do

What I would tell you to do

5 Claude Skills for LinkedIn

1. Viral Post Breakdown

Skill Name

Skill Description

Project Instructions

2. Content Auditor

Skill Name

Skill Description

Project Instructions

3. Content Repurposer

Skill Name

Skill Description

Project Instructions

4. Post Reviewer

Skill Name

Skill Description

Project Instructions

5. LinkedIn Carousel Generator

Skill Name

Skill Description

Claude Skill that creates LinkedIn carousels in Minutes

Skill details (copy and paste these)Skill Name

Skill Description

Project Instructions

The 3-tool cold email stack that works in 2026

The stack

Step 1 - - Research with Grok

Step 2 - - Write with Claude

Step 3 - - Send at scale with Maildoso

The full workflow in 60 seconds

Test Mailodoso now

Bonus -- My personal golden sequence

The 360° LinkedIn Profile & Content Audit Prompt

What it Covers

Profile Side:

Content side

Turn YouTube Videos Into LinkedIn Carousels

The 5-Step Process

Step 1 - Find the right video

Step 2 - Extract the key ideas

Step 3 - Turn ideas into a slide outline

Step 4 - Generate the visuals

Step 5 - Final polish in Canva

Claude Co-Work Workflows for LinkedIn

Workflow 1: Make your posts searchable

Workflow 2: Turn viral posts into repeatable formats

Workflow 3: Turn one idea into five posts

Workflow 4: Turn conversations into content

Workflow 5: Mine long content for posts

LinkedIn profile Audit System (PROMPT)

Why most LinkedIn profiles fail in under 10 seconds

Grok + LinkedIn

Prompt 1: Find ideas people are already arguing about

Prompt 2: Knowing the objectives before you post

Prompt 3: Validate whether an idea has signal or is already exhausted

Prompt 4: See how problems show up in conversations

Prompt 5: Say what people think but won’t post

SDRs: Never Start a Month at Zero

SDRs wait for the new month to begin before they start booking meetings. That habit kills the possibility to overachieve concistently.

Test Mailodoso now

HERE IS THE COMPLETE 3 EMAIL SEQUENCE

Email 1

Email 2

Email 3

Skill details (copy and paste these)

Skill Name