austin ai clubThe hub
№ 014marketing · content engine filed may '26

The skill graph.

One folder. Seventeen markdown files. Ten social accounts. $8K/mo → $0.

Austin AI Guy·22 min read·one weekend to build

We run ten social accounts and don't write a single post manually. No content team. No $8–12K/mo agency retainer. No sitting in front of Claude rewriting the same post ten times for ten platforms.

Just a folder of .md files, one AI agent, and a system that takes one idea and produces ten platform-native posts — each one actually thinking about the topic differently.

This is the full breakdown. Every file. Every wikilink. Every node. By the end you'll have a complete content production system you can realistically build this weekend.

01

The concept.

Most people use AI for content like this: open Claude. Type "write me a LinkedIn post about productivity." Get a generic post that sounds like a corporate intern wrote it. Spend 20 minutes making it not sound like a robot. Then do the whole thing again for every platform.

That's not a system. It's a chore with extra steps.

The problem isn't the AI — it's that you're giving it zero context about your brand, audience, voice, platform strategies, or how any of it connects. You're hiring a genius with amnesia every time you start a new chat.

Technically: a folder of interconnected markdown files where each file is one knowledge node — one piece of your content system's brain. Inside each file you use [[wikilinks]] (double-bracket references like [[brand-voice]] or [[hooks]]) to point to other nodes.

When you point an AI agent at this folder and give it a topic, it doesn't just read one file. It follows the links, reads the connected nodes, and builds up a complete understanding of your brand, voice, audience, platform rules, hook formulas, and repurposing logic — before writing a single word.

One flat file gives you a tool. A graph gives you a team — with sub-specialists for every platform, hook type, voice variant, and audience segment.

02

The stack.

  • 01Obsidian — free markdown editor with native [[wikilink]] support + graph viewdaily
  • 02Claude Projects — uploads all 17 files as persistent contextdaily
  • 03Plain folder (alt) — works just as well; the AI doesn't care about Obsidianfine
  • 04Claude Code or Cursor (advanced) — agent reads files directly from disk and can update thempowerful
03

The folder structure.

17 files. 4 folders. Your entire content production machine.

/content-skill-graph
├── index.md
├── platforms/
│   ├── x.md
│   ├── linkedin.md
│   ├── instagram.md
│   ├── tiktok.md
│   ├── youtube.md
│   ├── threads.md
│   ├── facebook.md
│   └── newsletter.md
├── voice/
│   ├── brand-voice.md
│   └── platform-tone.md
├── engine/
│   ├── hooks.md
│   ├── repurpose.md
│   ├── scheduling.md
│   └── content-types.md
└── audience/
    ├── builders.md
    └── casual.md
platforms/

→ native

One file per platform. Rules, formats, character limits, posting frequency. Everything the agent needs to write natively for that platform.

voice/

→ DNA

Brand voice + how it adapts per platform. The reason your content stops sounding like a robot generated it.

engine/

→ system

Hook formulas. The repurposing chain. Scheduling. Format definitions. The operational backbone that makes 1 idea → 10 posts work.

04

What the output actually looks like.

This is where most people mess up: rethinking, not reformatting. Same topic — How I use AI to manage 10 social accounts — eight different angles.

  • xX — "you don't need a content team. you need 30 markdown files. here's how i run 10 accounts without writing a single post manually:"
  • inLinkedIn — "6 months ago, I was spending $8,000 a month on content production across 10 platforms. Today I spend $0. Here's exactly what changed."
  • igInstagram — "I Run 10 Accounts And Don't Write Anything. (Swipe →)"
  • ttTikTok — "You're still writing content manually for every single platform? Let me show you what I use instead."
  • ytYouTube — "How to Run 10 Social Media Accounts with AI (Complete System Walkthrough)"
  • nlNewsletter — "This week I want to pull back the curtain on something I've been building quietly for months…"
  • thThreads — "hot take: the future of content isn't AI that writes for you. it's AI that thinks like 10 different people for you."
  • fbFacebook — "Has anyone else tried building a system to manage multiple social accounts at once? Curious what you all think."
05

The take.

A skill graph isn't a productivity hack. It's a transfer — everything you know about your content system, encoded once, executed forever. The agent gets smarter every week you spend ten minutes updating a node.

Steal one thing: start with index.md and brand-voice.md. Those two files define everything else. Build them well and the other 15 mostly write themselves.

Decision rules three signals worth acting on
If you see… Do this Don't do this
All 8 platform posts feel like clones Expand platform-tone.md with concrete example sentences per platform. Add the "rethink, don't reformat" rule to your starter prompt every run. Switch AI models.
Outputs sound like an AI wrote them Triple the "never use" list in brand-voice.md. Add 5 anchoring sample posts written by you. Voice is taught by example, not description. Add more adjectives to the brand-voice description.
Folder grew past 25 files and outputs got worse Audit for redundancy. Two files saying contradictory things about voice = the agent compromises. Pick one source of truth per concept. Add a 26th file to clarify.
Pairs with№ 001 — newsletter operating system →
Next in the libraryBrowse all 12 guides →

Want this wired into your business — not built from scratch this weekend? That's what the audit + retainer relationship is for. austinaiguy.com.