x-user-skill-creator

Generate an AI agent skill from any X/Twitter user's tweets.

Give a Twitter profile URL → auto-scrape tweets → distill knowledge → generate a ready-to-install skill.

Example output: navalskill — Naval Ravikant, 2,347 tweets → 1 skill with 6 sections + 9 knowledge packs + atomic library.

---

Install

npx skills add EwingYangs/x-user-skill-creator

Usage

In Claude Code:

/x-user-skill-creator https://x.com/paulg

Or:

Help me generate a skill from @elonmusk's tweets

What it does

Input: Twitter profile URL (e.g. https://x.com/naval)
  ↓
Step 0: Choose scrape scale (500 / 1000 / 3000 / 3000+ / custom)
Step 1: Scrape tweets (browser-use + date-range search)
Step 2: Filter & clean (remove replies/retweets, deduplicate)
Step 3: Discover topics → ★ human review (confirm/adjust topics)
Step 4: Build atomic library (keyword matching with confirmed topics)
Step 5: Generate knowledge packs (by topic, cross-section shared)
Step 6: Extract axioms & speaking style → ★ human review
Step 7: Generate SKILL.md (router + sections + workflows)
Step 8: Package for install
  ↓
Output: Ready-to-install skill directory

Output structure

产出目录结构参考 dontbesilent/dbskill：SKILL.md 作为主入口 + references/ 知识包 + 原子库。

{handle}/
├── SKILL.md                    # Router + all sections
├── references/                 # Knowledge packs (by topic)
│   ├── {topic_1}.md
│   ├── {topic_2}.md
│   └── ...
├── 原子库/
│   ├── atoms.jsonl             # Structured knowledge atoms
│   └── README.md
└── 高频概念词典.md              # High-frequency concept dictionary

Human Review

Two review checkpoints where you confirm before proceeding:

★ Step 3 — Topic discovery: Claude proposes 6-10 topics from tweet samples. You confirm, merge, split, or adjust before any classification runs.

★ Step 6 — Axioms & style: Claude extracts core axioms and speaking style per topic. You verify they accurately represent this person's philosophy before the skill is generated.

Scripts

Reusable scripts in x-user-skill-creator/scripts/:

Script	What it does
scrape_tweets.py	Scrape tweets via browser-use + date-range search
build_atoms.py	Convert tweets to structured knowledge atoms
build_knowledge_packs.py	Generate topic-based reference docs + frequency dictionary

Requirements

• Chrome browser logged into X/Twitter
• browser-use CLI (auto-detected; install if missing):

  curl -fsSL https://browser-use.com/cli/install.sh | bash

• python3

Related

• Twitter2Notion — Auto-sync your Twitter Bookmarks & My Tweets to Notion.