In a recent LinkedIn post, Ruben Hassid offers a detailed, step-by-step guide for leveraging AI, specifically Claude, to significantly improve LinkedIn content creation. Hassid argues that instead of generic prompts, users should develop personalized AI skills based on their own past performance. He outlines a process designed to move beyond simple AI assistance to a more sophisticated, data-driven approach that refines a user’s unique voice and content strategy.
From Scraping to Skill: Hassid’s AI Content Workflow
Hassid’s methodology centers on analyzing an individual’s historical LinkedIn posts to train an AI model. He emphasizes that the most effective AI-generated content stems from understanding what has already resonated with an audience. This approach contrasts sharply with common practices of asking AI to generate content from scratch.
The process begins with data acquisition. Hassid suggests using a tool like Apify to scrape existing LinkedIn posts.
“Stop prompting ‘Write a LinkedIn post for me.’ You only need one Claude skill (download for free).”
He explains that scraping one’s own posts provides the raw material for training a personalized AI model. According to Hassid, this initial data collection phase is crucial for establishing a foundation based on proven success.
Developing a Personalized AI Skill
The core of Hassid’s strategy involves using Claude Cowork, rather than the standard chat interface, to process the scraped data. By uploading an Excel file of past posts and providing specific prompts, users can instruct Claude to analyze their content for patterns and effectiveness.
Hassid highlights the depth of this analysis:
“It works with agents across all 400+ rows in Excel. It opens your images & carousels, not just text. Mine ran 10 minutes. Let it cook.”
The output of this process, as Hassid describes, is a comprehensive report detailing what elements of past posts were successful (hook, format, angle) and a Standard Operating Procedure (SOP) that codifies the recipe for creating high-performing content. This SOP then becomes the basis for creating a custom AI skill within Claude.
Transforming Data into a Custom AI Tool
The final stages of Hassid’s method involve converting the insights from the report and SOP into a functional AI skill. He provides a specific prompt structure for Claude’s `/skill-creator` tool, emphasizing that the skill should be named personally (e.g., “yourname-viral”) and designed to elicit necessary information from the user before generating content options.
This ensures the AI doesn’t just generate generic content but actively engages the user to gather context relevant to their specific needs, mirroring the user’s own thought process.
“Since you can’t create a viral post without having input for me, when I invoke the skill, you must use the AskUserQuestion tool, tailored to what Grant does best, to get the right input before generating angles/hooks/captions.”
Hassid reassures potential users who might feel this process is too complex or that their voice is too unique to be replicated by AI. He argues:
“But your voice isn’t a mystery. It’s a pattern you’ve already written 400 times.”
The continuous improvement loop is also a key aspect. Hassid advocates for regularly updating the AI skill with new successful posts to ensure it evolves with the user’s changing content strategy and audience engagement. This iterative process, he suggests, leads to increasingly effective content generation over time.
📝 About This Content
This article is based on insights shared by Ruben Hassid on LinkedIn.
📅 Originally posted on June 22, 2026 | View original post on LinkedIn →