Difficulty: Beginner Time to set up: 15 minutes Daily use: 5 minutes

Build a Claude Code skill that generates LinkedIn posts in your brand voice. 15-minute setup, 5-minute daily use.

---

Why This Works

Most people open ChatGPT, type "write me a LinkedIn post about AI," and get back something that sounds like a corporate press release. The problem is not the AI. The problem is that you gave it zero context about who you are, how you write, or what your audience cares about.

Claude Code fixes this by letting you build persistent context files that live on your machine. You set up your voice rules once, feed it examples of your best writing, and then every post it generates sounds like you wrote it on a good day.

This is the "Voice DNA" approach that blew up on Reddit a few months back. The idea is simple: instead of prompting from scratch every time, you encode your writing DNA into files that Claude Code reads automatically.

What You Will Build

A skill file that generates LinkedIn posts matching your exact tone, structure, and style. You will feed it your best posts as examples, define your rules, and then generate new posts with a single command.

Tools used: Claude Code (terminal), CLAUDE.md, skill files

---

Step 1: Install Claude Code

If you already have Claude Code running, skip to Step 2.

Open your terminal and run:

npm install -g @anthropic-ai/claude-code

Set your API key:

export ANTHROPIC_API_KEY=your-key-here

Add that export line to your ~/.zshrc or ~/.bashrc so it persists across sessions.

Test it works:

claude "Say hello"

You should get a response. If you do, you are good to go.

Step 2: Create Your Project Folder and CLAUDE.md

Create a dedicated folder for your marketing system:

mkdir -p ~/marketing-system/skills
cd ~/marketing-system

Now create your CLAUDE.md file. This is the file Claude Code reads automatically every time you start a session in this directory. Think of it as the brain that tells Claude who you are.

Create CLAUDE.md with the following structure:

# Marketing System

## About Me
- Name: [Your name]
- Role: [Your role, e.g., "Founder of a B2B SaaS company"]
- Industry: [Your industry]
- LinkedIn audience: [Who follows you, e.g., "marketing leaders, founders, AI practitioners"]

## Brand Voice Rules
- Write like a practitioner, not a thought leader
- Use short sentences. Punch hard.
- First person always
- No corporate jargon: say "we tested" not "we leveraged synergies"
- Contrarian takes welcome. Agreeable fluff is not.
- Specific numbers beat vague claims

## Skill Files
- `skills/linkedin-post.md` — LinkedIn post generator

This file does not need to be long. It needs to be accurate. The more specific you are about your voice, the better the output.

Step 3: Build the LinkedIn Post Skill File

Create skills/linkedin-post.md:

# LinkedIn Post Generator

## Post Structure
Every LinkedIn post follows this structure:

1. **Hook (line 1-2):** A bold claim, surprising stat, or contrarian take.
   Must stop the scroll. No "I'm excited to announce" garbage.
2. **Body (lines 3-15):** The substance. Story, framework, or proof.
   Use line breaks between thoughts. Short paragraphs only.
3. **CTA (last 2 lines):** Ask a question or invite discussion.
   Not "Like and share!" but something that invites real replies.

## Formatting Rules
- Max 1,300 characters (sweet spot for engagement)
- One sentence per line for readability
- No hashtags in the body. Max 3 hashtags at the very end, separated by a blank line.
- No links in the post body (kills reach). Put links in comments.
- Use "I" not "we" unless specifically about a team effort

## Tone Rules
- Confident but not arrogant
- Teach from experience, not theory
- Include at least one specific detail (number, name, date, result)
- If you would not say it in a conversation with a smart peer, do not write it

## Hook Templates That Work
- "Most [audience] think [common belief]. Here's what actually works:"
- "[Specific result] in [timeframe]. Here's the system:"
- "I stopped doing [common practice] 6 months ago. Here's what happened:"
- "The [industry] advice no one gives you:"
- "I [did something unexpected]. [One line of context]."

## What NOT To Do
- No "I'm thrilled to share"
- No walls of text with no line breaks
- No vague advice without examples
- No humble brags disguised as lessons
- No posts that could have been written by anyone

Step 4: Feed It Your Best Posts as Examples

This is the step most people skip, and it is the most important one. Create a file called skills/linkedin-examples.md:

# My Best LinkedIn Posts (Examples)

Use these as style and tone references. Match the voice, structure,
and energy of these posts when generating new ones.

## Example 1
[Paste your highest-performing LinkedIn post here, exactly as published]

## Example 2
[Paste another one]

## Example 3
[Another one]

Aim for 5 to 10 examples. Pick posts that got good engagement AND that you are proud of. Not every viral post represents your best writing. Choose the ones where you think "yes, that sounds like me at my best."

If you do not have 10 good posts yet, start with 3 to 5. You can add more over time.

The Reddit "Voice DNA" thread nailed this: the examples are more powerful than the rules. Rules tell the AI what to do. Examples show it how you actually do it. Both together are what make the output feel like you.

Step 5: Generate Posts

Now you are ready. Open Claude Code in your marketing-system folder:

cd ~/marketing-system
claude

Then give it a topic:

Write a LinkedIn post about why most companies are invisible to AI search engines

Claude Code will automatically read your CLAUDE.md, which points to your skill files and examples. The output should sound like your voice, follow your structure rules, and hit the right length.

Here are more prompt examples you can use:

Write a LinkedIn post about a specific win we had with content repurposing last week.
The win: turned one blog post into 14 pieces of content that drove 3x our normal traffic.

Write a LinkedIn post sharing a contrarian take: most marketing teams
should fire their SEO agency and invest in AI visibility instead.

Write a LinkedIn post teaching a simple framework.
Topic: How to audit whether your website is visible to ChatGPT and Perplexity.
Make it a numbered list post.

You can also ask for variations:

Give me 3 different hooks for a post about AI-powered content systems.
Then write the full post using the best one.

Step 6: Edit and Publish

The AI drafts. You refine. This is not a "set it and forget it" system. Every post should go through your human filter before it goes live.

Here is the editing checklist:

1. Read it out loud. Does it sound like you talking to a friend? If it sounds like a blog post, tighten it up.
2. Check the hook. Would you stop scrolling for this? If not, rewrite line one.
3. Cut the filler. If a sentence does not add information or emotion, delete it.
4. Add a specific detail. If the post feels generic, add one real number, name, or date from your experience.
5. Test the CTA. Would you actually reply to this question? If not, ask a better one.

The whole process from prompt to published post should take about 5 minutes once your skill files are dialed in.

File Structure Recap

Here is what your folder should look like when you are done:

marketing-system/
  CLAUDE.md
  skills/
    linkedin-post.md
    linkedin-examples.md

That is it. Three files. You now have a LinkedIn post generator that sounds like you, follows your rules, and gets better as you add more examples over time.

Making It Better Over Time

Every time you publish a post that performs well, add it to your examples file. Every time the AI produces something that feels off, add a new rule to your skill file about what to avoid.

This is a living system, not a static template. The marketers who get the most out of Claude Code treat these files like a garden: tend them regularly, and the output keeps improving.

---

Tools used in this playbook: Claude Code CLI, CLAUDE.md configuration, skill files, markdown