borghei/Claude-Skills / marketing/x-twitter-growth
x-twitter-growth
VerifiedThis skill should be used when the user asks to "analyze tweets", "grow on Twitter", "build Twitter threads", "optimize X posting schedule", "track follower growth", "improve tweet engagement", or "create a Twitter content strategy".
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 6.0 KB |
| examples/tweets.csv | other | 5.9 KB |
| references/x-growth-playbook.md | markdown | 6.4 KB |
| scripts/growth_tracker.py | script | 13.5 KB |
| scripts/thread_builder.py | script | 9.9 KB |
| scripts/tweet_analyzer.py | script | 14.4 KB |
references/x-growth-playbook.md
View on GitHub# X/Twitter Growth Playbook
## Algorithm Fundamentals
### How X Ranks Content
> **Historical vs. current (as of September 2026).** The candidate counts, 50/50 mix, and signal weights below come from X's **2023** open-source release of its recommendation algorithm. They are historical and approximate. In 2026 X published a newer For You ranker at [github.com/xai-org/x-algorithm](https://github.com/xai-org/x-algorithm): in-network and out-of-network posts are ranked together by a transformer model that predicts how likely *this viewer* is to take each action (like, reply, repost, dwell, block, report, etc.), and configurable weights blend those predicted probabilities into a score. The repo notes that the weights scale each viewer's *predicted* probability of an action, not raw engagement counts — so "one reply = N likes" arithmetic does not hold. Check the repo for current parameters before quoting numbers.
The 2023 system used a multi-stage ranking pipeline:
1. **Candidate generation** - Selects ~1,500 candidate tweets
2. **Feature extraction** - Analyzes tweet and author signals
3. **Ranking model** - Scores each candidate
4. **Filtering** - Removes low-quality, duplicate, blocked content
5. **Mixing** - Blends in-network (~50%) and out-of-network (~50%) content
The current ranker keeps the same shape (retrieve in- and out-of-network candidates, hydrate, filter, score, then blend in ads and recommendations), with visibility filtering decided separately from ranking.
### Key Ranking Signals (2023 release — historical)
The durable lesson holds in both versions: conversation (replies, especially replies you engage with), reposts, and dwell outweigh passive likes, and negative feedback (mutes, blocks, reports) is heavily penalized.
**Positive signals (ordered by weight in the 2023 release; multiples are approximate):**
1. Reply (weighted ~27x a like)
2. Retweet/Quote tweet (weighted ~4x a like)
3. Like
4. Bookmark
5. Time spent reading
6. Profile click after viewing
7. Link click
**Negative signals:**
- "Not interested" or mute
- Unfollow after viewing
- Report
- Block
### Content Type Multipliers (unofficial heuristics)
These are practitioner rules of thumb, not published X parameters — use them as testing hypotheses.
- Text-only tweets: 1.0x baseline
- Tweets with images: 2.0x
- Tweets with video: 2.5x (if watched >50%)
- Threads: Variable (hook tweet gets algorithm boost if thread performs)
- Quote tweets: 1.5x
- External links: 0.5x (algorithm deprioritizes link-outs)
## Content Format Frameworks
### The Hook-Value-CTA Framework
Every high-performing tweet follows this structure:
1. **Hook** (first line) - Stop the scroll, create curiosity
2. **Value** (body) - Deliver on the hook's promise
3. **CTA** (last line) - Drive specific engagement action
### Hook Formulas That Work
1. **Contrarian take:** "Most people think X. They're wrong. Here's why..."
2. **Number list:** "7 lessons from [experience] that changed my approach to..."
3. **Story opener:** "In 2024, I [did X]. It was the best decision I ever made. Here's why..."
4. **Question:** "Why do 90% of [people] fail at [thing]?"
5. **Bold claim:** "[Thing] is dead. Here's what's replacing it..."
6. **Personal result:** "I went from [A] to [B] in [time]. Here's exactly how..."
### Thread Writing Best Practices
**Structure:**
- Tweet 1: Hook (strongest possible - determines thread performance)
- Tweet 2-3: Context and setup
- Tweet 4-N: Value delivery (one idea per tweet)
- Final tweet: Summary + CTA + repost of tweet 1
**Thread rules:**
- 7-12 tweets is optimal length (enough value, not too long)
- Each tweet should work standalone (people scroll)
- Use line breaks for readability
- Number your tweets (1/, 2/, etc.) only if educational
- End each tweet with a reason to read the next one
**Thread engagement tactics:**
- Ask a question in tweet 3-4 to drive replies mid-thread
- Include one data point or statistic per thread
- Add an image in tweet 2 or 3 (breaks up text)
- Quote tweet your own thread as a summary 2-4 hours later
## Engagement Optimization
### Posting Time Optimization
**General best times (US audience):**
- Weekdays: 8-10 AM, 12-1 PM, 5-7 PM (EST)
- Weekends: 9-11 AM (EST)
- Worst times: 1-5 AM, late Friday evening
**Important:** Your audience's optimal times may differ. Use `growth_tracker.py` to identify your specific best times from actual data.
### Reply Strategy
Replying to others is the #1 growth lever for small accounts:
- Reply to larger accounts in your niche (first 30 minutes of their tweet)
- Add genuine value, not just agreement
- Your reply gets shown to their followers
- Aim for 10-20 thoughtful replies daily
### Engagement Pods vs Organic
Engagement pods (coordinated liking/replying) provide short-term boosts but:
- Algorithm detects artificial patterns
- Doesn't build real audience
- Creates engagement debt (metrics look good, growth is hollow)
Focus on organic engagement through quality content and genuine interactions.
## Growth Compounding Strategies
### The 1-3-5 Daily System
- **1** long-form tweet or thread (original content)
- **3** replies to accounts in your niche
- **5** engagement actions (likes, retweets, bookmarks)
### Content Recycling
Top-performing content can be recycled:
- Wait 60-90 days before reposting similar content
- Rephrase with fresh angle, not exact copy
- Turn threads into single tweets (and vice versa)
- Convert tweet insights into threads with expanded detail
### Follower Growth Stages
**0-1,000 followers:**
- Focus: Reply game, niche identification
- Content: 70% replies, 30% original
- Goal: 50-100 followers/month
**1,000-10,000 followers:**
- Focus: Consistent original content, threads
- Content: 50% original, 30% replies, 20% engagement
- Goal: 500-1,000 followers/month
**10,000-50,000 followers:**
- Focus: Threads, collaborations, cross-promotion
- Content: 60% original, 20% threads, 20% engagement
- Goal: 2,000-5,000 followers/month
**50,000+ followers:**
- Focus: Brand building, monetization, community
- Content: 70% original, 15% threads, 15% community
- Goal: Engagement rate maintenance over raw growth
### Profile Optimization Checklist
- Professional headshot or distinctive avatar
- Name: Real name + keyword (e.g., "Jane Smith | Growth Marketing")
- Bio: What you do + who you help + social proof (160 chars max)
- Pinned tweet: Best thread or signature content piece
- Banner: CTA or value proposition visual
- Link: Newsletter, website, or link tree
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |