LinkedIn Post Maker

A multi-agent Claude Code pipeline that turns a topic or rough draft into a polished, human-sounding LinkedIn post. It researches the topic, writes 3 variants, critiques them, humanizes them, and picks the best one — all automatically.

What it does

You give it a topic. It gives you a ready-to-post LinkedIn post.

Behind the scenes, 5 specialized AI agents run in sequence:

1. Researcher — fetches any URLs you provide, then searches the web for supporting stats, news, and industry examples
2. Writer — writes 3 post variants: short & punchy, story-led, and contrarian
3. Skeptic — critiques each variant for buzzwords, vague claims, and AI-sounding patterns
4. Humanizer — rewrites each variant using the critique to make it sound like a real person
5. Judge — scores all 3 humanized variants and picks the winner

The full session (research, variants, critiques, scorecard, winning post) is saved to posts/ as a markdown file.

Requirements

• Claude Code — the CLI tool from Anthropic

Setup

1. Clone or copy this directory to your local machine.

2. Customize config.md for your voice. Open config.md in the project root and update it with your details. This is the only file you need to change — the agents read it automatically.

config.md
├── About You       — your name, role, company, website
├── What You Do     — one-line description of your work
├── Your Audience   — who reads your posts and what they care about
├── Your Voice      — how you write: tone, style, what to avoid
└── Topics          — the subjects you post about

3. Open the project in Claude Code:

cd linkedin-post-maker
claude

Usage

Run the /linkedin-post command followed by your topic:

/linkedin-post why most AI pilots in banking never reach production

Pass a URL and the researcher will read it as the primary source:

/linkedin-post https://example.com/article-about-ai-in-banking

Or hand it a rough draft to refine:

/linkedin-post I want to write about how we helped a regional bank cut their loan review time in half using AI agents. Here's a rough draft: [paste draft]

The pipeline runs automatically. When it's done, you'll see:

• The winning post ready to copy and paste
• The runner-up in case you prefer it
• A one-line reason why the winner was chosen

The full session is saved to posts/YYYY-MM-DD-topic-slug.md.

Tips

• Be specific with your topic. "AI in banking" is too broad. "Why AI agents are better than RPA for back-office ops" gives the researcher something to work with.
• Pass a URL to anchor the post. If there's an article, report, or announcement you want to respond to, include the URL. The researcher reads it first and uses it as the primary source.
• Share real context if you have it. If you have a real story, stat, or client observation, include it in your prompt. The agents are instructed never to fabricate anecdotes — so real material makes the post stronger.
• Check the runner-up. Sometimes it's actually better for your specific situation.
• The full session file is useful. It contains the research brief, all 3 variants, and the critique — good raw material if you want to manually tweak the winning post.

Project structure

linkedin-post-maker/
├── .claude/
│   ├── agents/          # The 6 specialist agent definitions
│   └── commands/        # The /linkedin-post slash command
├── posts/               # Generated posts (one file per run)
├── config.md            # Your author config — edit this to customize the pipeline
└── CLAUDE.md            # Project instructions for Claude Code