Most people use AI for content the lazy way - throw in a topic, get generic output, post it anyway or get frustrated and write it themselves.

I wanted something better: a repeatable system where I point Claude Code at a video I recorded, and it gives me back a LinkedIn post and a blog post that actually sound like me.

So I built it. And I recorded the whole thing so you can build it yourself.

The Problem with AI-Generated Content

You already know this: telling Claude "write me a LinkedIn post about X" gives you generic output. It doesn't know your voice, your examples, your style.

But here's the thing — even if you have a second brain with all your context, you still need a repeatable system. Otherwise you're doing the same thing every time: finding the right files, giving instructions, formatting the output.

That's what this workflow solves.

What We're Building

A system that takes a video transcript and turns it into a LinkedIn post and a blog post — using your own content as the style reference.

The pieces:

  • Tella API → fetches video transcripts automatically

  • Apify LinkedIn scraper → scrapes your best-performing LinkedIn posts

  • Claude Code skills → reusable instructions that do the work every time

Once built, the workflow is simple: point it at a video, get a LinkedIn post and blog post back. In your voice, using your structure.

Step 1: Scrape Your LinkedIn Posts as Style References

This is the part most people skip. They want AI to write in their voice, but they never show it what their voice looks like.

I used an Apify actor (supreme_coder/linkedin-post) to scrape my last LinkedIn posts. Then I filtered to only keep the ones with 10+ likes — if they resonated with people, they're probably a better style reference.

The result: 35 posts saved as markdown files, each with the content, date, reaction count, and URL.

One important detail: not all posts work as references for all topics. A career change post has a completely different structure than a workflow tutorial. So the skill filters by topic — if I'm writing about an automation, it only loads posts about automations.

This is how the LinkedIn post folder looks like. We have all posts and the posts generated using Claude Code in a different folder.

Step 2: Connect the Tella API

I already had my Tella videos stored locally from a previous fetch. But videos get added constantly, and I don't want to re-run the full script every time.

So the skill has a fallback: check the local folder first, and if the video isn't there, fetch it directly from the Tella API. This way it always has access to the latest content, whether it's cached or not.

The API returns the full transcript with timestamps, chapters, and metadata — everything you need to understand what the video covers.

This is a screenshot of my Tella videos as I have them organized in the context folder.

Step 3: Build the Claude Code Skills

This is where it comes together. A skill in Claude Code is basically a saved prompt — a set of instructions that you can reuse without re-explaining everything.

I created two separate skills:

LinkedIn Post Skill:

  1. Load the video transcript (local or API)

  2. Load only the topic-relevant LinkedIn posts as style references

  3. Analyze the hook patterns, structure, and tone from those references

  4. Generate a post that matches — not a copy, but the same feel

Blog Post Skill:
Same idea, different output format. Blog posts are longer (800-1,500 words), have section headers, and include more detail. They use my vibesales.co posts as style references instead of LinkedIn.

Both skills pull from the same transcript but produce very different content — because the references are different.

Part of the LinkedIn post creation skill - if you want the entire skill let me know!

Step 4: Run Both Together

Once the skills exist, using them is straightforward:

"Help me create a blog post and a LinkedIn post using my automation with Claude Code and n8n video."

Claude Code reads the transcript, fetches the right style references for each output, and generates both. The LinkedIn post is short, punchy, hook-driven. The blog post is detailed, scannable, tutorial-style.

Same source content. Different formats. Both in my voice.

If you want to see the end results:

Taking It Further: Trigger-Based Automation

Right now this is manual — I tell Claude Code which video to use, and it runs the skills. But you can take it further.

With n8n or Trigger.dev, you can set up a trigger-based workflow:

  1. Trigger: New Tella video detected (check daily or weekly)

  2. Action: Fetch the transcript

  3. Action: Run the LinkedIn post and blog post skills

  4. Action: Send drafts to Slack or Google Drive for review

At that point, every video you record automatically produces content drafts. You just review and publish.

I haven't built this yet, but the foundation is there. The skills already have the logic — you just need a trigger to kick them off.

Why This Matters

This isn't about saving time (although it does). It's about consistency.

When you write manually, some posts are great and some are rushed. When you use AI without context, everything sounds the same — just not like you.

This system gives you both: the speed of AI with the authenticity of your own content. And because the skills are reusable, the quality stays consistent every time.

Getting Started

If you want to build this yourself:

  1. Start with your style references — scrape or save your best posts

  2. Get your content source — Tella, Fathom, any tool with transcripts

  3. Build one skill — start with LinkedIn posts, they're shorter

  4. Test and refine — the first output won't be perfect, tweak the instructions

  5. Add more skills — blog posts, YouTube descriptions, email sequences

The full step-by-step build is in the video. I show the entire process from zero, including the mistakes and adjustments along the way.

What would you automate first — LinkedIn posts, blog posts, or something else entirely? I'm curious what content takes you the most time.

Remember: Choose Vibes instead of stress, and watch your Sales grow!

Until next time,

Santiago 😎

Keep Reading

No posts found