AI Content Pipeline

Overview

Keeping up with the cybersecurity news cycle and maintaining a consistent LinkedIn presence at the same time is a lot to ask. Reading through RSS feeds, deciding what is worth sharing, writing a post that sounds like you, generating a matching image, and getting it all organized. That is a full workflow in itself. This pipeline automates every step of it.

Built on n8n, the pipeline subscribes to a curated set of cybersecurity RSS feeds. Whenever a new article is published, the workflow triggers automatically. An AI agent reads the article and decides whether it is worth sharing (filtering out webinar announcements, product promotions, and sponsored content), then writes a LinkedIn post in Hebrew, following a specific tone and content standard.

After the post is generated, a dedicated sub-workflow creates a matching image using Freepik’s Flux model, uploads it to GitHub for hosting, and logs the complete package to a Google Sheet: the Hebrew post, an English summary of the article, and the image URL, ready to review and publish.

Architecture

RSS Monitoring

The pipeline subscribes to RSS feeds from major cybersecurity news sources. Each new article fires the workflow automatically, without polling on a schedule. The trigger passes the article URL and metadata downstream for processing.

Content Filtering

An AI agent reads each incoming article and classifies it before anything else runs. Webinar announcements, executive interviews, product launches, and sponsored content are discarded. Only articles covering actual security events (vulnerabilities, active exploits, threat actor campaigns, or meaningful research) pass through to the next stage. This keeps the pipeline focused and the output consistent.

Post Generation

Articles that pass the filter are handed to the AI agent for content creation. The agent writes a LinkedIn post in Hebrew, matching a defined personal tone and structure. Alongside the Hebrew post, the agent produces a concise English summary of the source article. Both outputs are structured and ready for publishing without manual editing.

Image Generation

Once the post is ready, the main workflow calls a dedicated sub-workflow to handle the visual. The sub-workflow uses the post content as context to generate an image with Freepik’s Flux model, producing a visual that fits the topic and mood of the article. The generated image is then uploaded to GitHub for reliable public hosting, and the resulting URL is passed back to the main workflow.

Output & Logging

With the post and image URL in hand, the final step writes a new row to a dedicated Google Sheet. Each row contains the Hebrew LinkedIn post, the English article summary, and the GitHub-hosted image URL. The sheet serves as both an archive and a staging area, where content is reviewed before being published to LinkedIn.

Workflow Example

1. A new article appears in one of the monitored feeds: a critical zero-day vulnerability actively exploited in enterprise VPN software
2. The AI agent reads the article and classifies it as relevant security news (not a product announcement or recap)
3. The agent writes a Hebrew LinkedIn post covering the vulnerability, its impact, and what organizations should do
4. The agent produces a concise English summary of the original article
5. The image generation sub-workflow receives the post content and creates a matching visual using the Freepik Flux model
6. The generated image is uploaded to GitHub and the public URL is captured
7. The Hebrew post, English summary, and image URL are written as a new row in the Google Sheet
8. The post is reviewed and published to LinkedIn directly from the sheet