LinkedIn Algorithm Skill for Claude Code

Score and optimize LinkedIn posts using reverse-engineered algorithm signals from LinkedIn's production feed system (2026).

What's Included

• 5-Step Optimization Workflow: Format selection, draft analysis, algorithm optimization, scoring, rewrite suggestions
• Weighted Scoring System: Dwell (30%), Conversation (25%), Save/Share (20%), Format (15%), Safety (10%)
• Confidence Tags: Every claim tagged as CONFIRMED, LARGE-SCALE DATA, or PRACTITIONER ESTIMATE
• 2 Reference Files: Deep-dive into ranking signals and technical architecture

Research Sources

Built from primary sources, not recycled blog advice:

• 5 peer-reviewed arXiv papers on LinkedIn's feed ranking (360Brew, LiGR, Feed SR)
• LinkedIn VP/Director statements on hashtags, links, engagement detection
• van der Blom 1.8M post dataset
• AuthoredUp 3M+ post dataset
• Buffer 52M post dataset

Installation

1. Download SKILL.md and the references folder
2. Drag and drop into Claude (Desktop app, VS Code, Cursor, or Terminal)
3. Tell Claude: "Install this skill"
4. Invoke with /linkedin-algorithm-skill

Usage

/linkedin-algorithm-skill

Then ask Claude anything:

• "Score this LinkedIn post"
• "Optimize my LinkedIn post for maximum reach"
• "Why did my LinkedIn post flop?"
• "Write a LinkedIn post about [topic] optimized for the algorithm"
• "Should this be a carousel or a text post?"

How It Works

LinkedIn's feed is a 4-stage neural pipeline:

1. Retrieval: LLM-based dual encoder narrows candidates to ~2,000
2. Ranking: Sequential transformer scores using your last 1,000 interactions
3. Re-ranking: Attention-based diversity prevents repetitive feeds
4. Serving: Trust classifiers and language matching applied

This skill optimizes for the signals these systems actually measure: dwell time, quality engagement, saves/shares, format fit, and negative signal avoidance.

Scoring Example

Category	Score	Weight
Dwell potential	8	30%
Conversation potential	7	25%
Save/share potential	6	20%
Format optimization	9	15%
Safety (no bait/pods/links)	10	10%

Weighted Score: 7.60 (Strong)

Key Insights

• Carousels (PDF documents) outperform all other formats across every dataset
• The "see more" click is a confirmed engagement signal
• 15+ word comments are valued more than "great post"
• 6+ hashtags correlates with significant reach drops
• Link posts underperform all native formats (put links in comments)
• LinkedIn refreshes embeddings every 30 minutes, making the first 60-90 minutes critical

License

Free to use and modify.

Author

Created by Attainment