The Claude LinkedIn Content System

Before you touch any automation, read this

Most people building AI content systems make the same mistake: they spend 80% of their time configuring Claude, connecting APIs, and building workflows - and produce nothing worth putting out.

The whole point of content is to get clients. If the output you are automating is mediocre, you have just built a machine that produces mediocre content faster.

The three rules that actually matter

Rule 1: Only automate what you can already do manually

Rule 2: Do the process manually at least once before automating it

Rule 3: The output test

The content machine architecture

The sequence that determines whether content grows your business:

→ Brand positioning: what makes your content different from everyone else in your space

→ Packaging: your idea, thumbnail, and title - in that order, because that is how people decide whether to click

→ Script: write it, film it

→ Editing

→ Analytics: review what worked, what did not, repeat

Packaging is 70-80% of success on YouTube. If nobody clicks, nobody watches. Claude helps you with the idea and packaging stage more than anywhere else.

Claude is not the creative director - you are

Claude can analyse data, spot patterns, generate options, and produce drafts. What it cannot do is make a judgment call on what is interesting, what is authentic to your voice, and what will resonate with your specific audience.

That human judgment is the competitive moat. The moment everyone can automate the same output with the same prompts, the output stops being differentiated. The people who win are the ones using Claude to save time on the tedious parts while keeping the creative decisions to themselves.

What this skill does

Instead of guessing what to make content about, this skill researches what your ICP is actually talking about across Reddit, X/Twitter, YouTube, Google Trends, and any other forums you specify - and surfaces high-demand, low-supply gaps you can fill.

The output tells you: what topics are trending, what everyone is saying that you could say the opposite of, and what content already exists so you do not duplicate it.

What to ask it to find

The four things to extract from the intelligence report

The view velocity signal

How to build this skill

You do not need to download someone else's skill for this. Build your own so it is specific to your niche.

Step 1: Use the built-in skill creator

Step 2: Run the research prompt

Step 3: Use the output to validate, not just generate

Thumbnails

How to use AI for thumbnails without just copying

Two thumbnail approaches

What AI is actually useful for in thumbnails

Titles

How to generate and test titles

Build a title skill for consistency

What to use the skill for

A YouTube script writer skill takes your raw idea and angle and produces:

→ An intro (showing the desired outcome, the cost of doing nothing, why now)

→ A chapter-by-chapter structure with main talking points

→ A mid-video CTA (typically around 3-4 minutes in)

→ An outro and close

The skill handles the structure. You handle the voice.

How to build the script writer skill

Use the built-in skill creator: type /skill and skill creator. Tell it: "I want a YouTube script writer skill. When I give it a video idea and angle, it should produce an intro, chapter outlines with talking points, a mid-video CTA, and an outro."

Then give it examples of scripts or videos you consider well-structured so it can calibrate to your style.

How to actually use the output

Do not copy and paste the script and read it verbatim. The output is a starting point, not the finished product.

The intro: rewrite it by hand

The body: use it as talking points

Why this is the most powerful use of Claude for content

Every piece of research you do about your ICP online - Reddit threads, YouTube comments, competitor teardowns - is someone else's interpretation of what your buyer wants. It is filtered and indirect.

Your own sales calls are direct, unfiltered, and impossible to replicate. Nobody else has had your exact conversations with your exact clients. That is the content moat.

When you run sales calls through a Claude skill, you extract the exact language, pains, desires, and blockers your buyers use - and turn that into content that speaks directly to them.

The workflow

Step 1: Record every call

Step 2: Build a sales call analyzer skill

Step 3: Feed in transcripts and extract content intelligence

Step 4: Build content directly from the output

Run this regularly

Do not run the sales call analyzer once and file the output. Re-run it after every 5-10 new calls. Buyer language shifts over time. New patterns emerge. The content topics that resonated 6 months ago may not be the highest-priority ones now.

If you are actively booking calls through prosp.ai outreach, you will have a steady stream of fresh data to feed into the skill. The more volume you run, the better your content intelligence becomes.

Automate this

Content market intelligence research

Script structure and talking points

Title variations

Sales call analysis

Basic video editing for talking-head formats

Keep humans here

The intro rewrite

Creative direction on thumbnails

Video editing when differentiation matters

The final judgment call on everything