🧊 IceCube Content Factory

Content that captures attention. Automatically.

Most content gets ignored. IceCube Content Factory uses engagement psychology to create content that stops the scroll.

What This Skill Does

1. Hook Generation

Generate 10 different hook styles for any topic:

• Pattern interrupt hooks
• Curiosity gap hooks
• Pain point hooks
• Story hooks
• Controversy hooks
• Social proof hooks
• FOMO hooks
• Counterintuitive hooks
• How-to hooks
• List hooks

2. Thread/Post Structures

• Twitter/X threads (optimal structure)
• LinkedIn posts (professional tone)
• Xiaohongshu notes (visual + emotional)
• Reddit posts (authentic + value)
• Blog intros (SEO + engagement)

3. Engagement Optimization

• Optimal posting times by platform
• Hashtag strategies
• CTA placement
• Emoji usage patterns
• Formatting for readability

4. Content Remixing

• Turn one idea into 10 posts
• Repurpose long-form into short-form
• Transform text into visual concepts
• Create content series from single topic

Hook Templates

1. Pattern Interrupt

"Everyone thinks [common belief].
Here's why they're wrong:"

2. Curiosity Gap

"I discovered something that changed everything.
Most people will never know this:"

3. Pain Point

"Struggling with [pain point]?
I spent 6 months figuring this out so you don't have to:"

4. Story

"6 months ago, I was [bad situation].
Today, [good outcome].
Here's exactly what changed:"

5. Controversy

"Unpopular opinion: [controversial take]
Let me explain:"

6. Social Proof

"[X people] have used this to [result].
Here's the breakdown:"

7. FOMO

"This [opportunity] is disappearing.
Those who act now will [benefit]:"

8. Counterintuitive

"The best way to [goal] is NOT [expected method].
It's actually [surprising method]:"

9. How-to

"How to [achieve outcome] in [timeframe]:
A step-by-step guide:"

10. List

"[Number] [things] that will [outcome]:
[Teaser 1]
[Teaser 2]
[Teaser 3]
Thread 🧵"

Usage Examples

Example 1: Twitter Thread

Input: "AI agent memory"
Output:
Hook: "Most AI agents forget everything after 30 minutes.
Here's how I built one that remembers forever:"

Structure:
1/8 [Hook]

2/8 The problem: Zep uses 600K tokens per conversation

3/8 The solution: File-based memory

4/8 How it works: Four-layer architecture

5/8 The results: 15KB vs 600KB

6/8 Implementation details

7/8 Lessons learned

8/8 If you're building AI agents, this matters.
RT if helpful 🔁

Example 2: Xiaohongshu Note

Input: "AI agent memory"
Output:
Title: "AI agent 记忆力提升 4000%｜从 600KB 降到 15KB"

Hook: "大部分人不知道 AI agent 为什么总是忘记..."

Structure:
- Problem (痛点)
- Solution (解决方案)
- Results (效果)
- How-to (教程)
- CTA (互动)

Emojis: 🧊 💾 ⚡ 📉

Hashtags: #AI工具 #效率提升 #黑科技

Example 3: LinkedIn Post

Input: "AI agent memory"
Output:
Hook: "After 6 months of experimentation, I finally solved the AI memory problem."

Body:
- Professional framing
- Data-driven results
- Business implications
- Call to action

Tone: Thought leadership, not clickbait

Content Psychology Principles

1. Open Loops

• Start with incomplete information
• Promise resolution at the end
• Keep readers engaged throughout

2. Pattern Interrupts

• Break expected patterns
• Surprise the brain
• Force attention

3. Value Density

• Every sentence must add value
• Cut fluff ruthlessly
• Respect reader time

4. Emotional Triggers

• Curiosity
• Fear of missing out
• Desire for gain
• Pain avoidance
• Social validation

5. Authority Signals

• Data and metrics
• Personal experience
• Expert quotes
• Case studies

Workflow

Step 1: Topic Input

topic: "AI agent memory"
audience: "developers"
platform: "twitter"
tone: "technical but accessible"
goal: "educate and drive interest"

Step 2: Hook Generation

Generate 5 hooks using different templates.

Step 3: Structure Selection

Choose optimal structure for platform.

Step 4: Content Drafting

Draft complete content with hooks + body + CTA.

Step 5: Optimization

• Check readability score
• Verify engagement elements
• Add platform-specific elements

Step 6: Output

Generate final content ready to post.

Advanced Features

A/B Test Hooks

Generate multiple hooks, test engagement:

hooks:
  - hook_a: "Most AI agents forget everything..."
    style: "pattern_interrupt"
  - hook_b: "I spent 6 months solving memory..."
    style: "story"
  - hook_c: "600KB vs 15KB: The memory breakthrough..."
    style: "counterintuitive"

Content Series

Turn one topic into a week of content:

Day 1: Hook-focused intro
Day 2: Deep dive #1
Day 3: Case study
Day 4: Deep dive #2
Day 5: Implementation guide
Day 6: Common mistakes
Day 7: Summary + CTA

Platform Optimization

twitter:
  max_chars: 280
  optimal_threads: 8-12 tweets
  hashtag_limit: 2-3

linkedin:
  max_chars: 3000
  optimal_length: 1300-2000
  no_hashtags_in_body

xiaohongshu:
  title_chars: 20
  body_style: emotional + visual
  emoji_density: high

reddit:
  max_title: 300
  style: authentic + value-dense
  no_obvious_promotion

Integration with IceCube Suite

icecube-memory: Store successful hooks and patterns

icecube-heartbeat: Track content performance during maintenance

icecube-evolution: Learn from high-engagement content

Output Format

memory/content/YYYY-MM-DD.md:

# Content Factory — YYYY-MM-DD

## Topic: AI Agent Memory
Platform: Twitter
Generated: HH:MM

### Hooks Generated
1. Pattern Interrupt: "Most AI agents forget..."
2. Story: "I spent 6 months..."
3. Counterintuitive: "600KB vs 15KB..."

### Selected Hook
"Most AI agents forget everything after 30 minutes.
Here's how I built one that remembers forever:"

### Full Content
[8-tweet thread]

### Engagement Prediction
- High: Hook type performs well in tech niche
- Risk: Might attract developer audience only

### Posted
- [ ] Twitter
- [ ] LinkedIn
- [ ] Xiaohongshu

### Actual Performance
(Updated after posting)
- Impressions: 
- Engagement:
- Click-through:

Anti-Patterns

❌ Don't:

• Use clickbait without substance
• Over-promise and under-deliver
• Ignore platform norms
• Copy-paste without adaptation

✅ Do:

• Deliver value in every piece
• Match hook to actual content
• Adapt tone per platform
• Learn from high performers

License

MIT — Use freely.

Content that stops the scroll. Not just more noise.