navalskill

Naval Ravikant wisdom toolkit. Built from 2,347 tweets (2010-2026), distilled into knowledge atoms, and packaged as an AI agent skill.

Inspired by dbskill — same architecture, different philosophy.

---

Install

npx skills add EwingYangs/navalskill

Or manually:

git clone https://github.com/EwingYangs/navalskill.git /tmp/navalskill && cp -r /tmp/navalskill ~/.claude/skills/navalskill && rm -rf /tmp/navalskill

After install, type /naval in Claude Code.

---

What it does

One skill (/naval) with 6 sub-sections, auto-routed by intent:

Sections cross-reference each other. For example:

• Wealth detects happiness issue → switches to Happiness
• Decision needs leverage analysis → switches to Leverage
• Health question is really anxiety → switches to Happiness

---

Skill Structure

navalskill/
├── SKILL.md                        # Router + 6 sub-sections
├── references/                     # Knowledge packs (by topic, cross-section)
│   ├── wealth_and_specific_knowledge.md
│   ├── leverage_and_scale.md
│   ├── happiness_and_peace.md
│   ├── decision_and_judgment.md
│   ├── learning_and_reading.md
│   ├── health_and_fitness.md
│   ├── philosophy_and_meaning.md
│   ├── relationships_and_trust.md
│   └── technology_and_startups.md
├── atoms/                           # Structured knowledge database
│   ├── atoms.jsonl                  # 2,347 knowledge atoms
│   └── README.md                    # Field docs
└── frequency_dictionary.md          # High-frequency concept dictionary

Knowledge packs are organized by topic (not by skill section), so multiple sections can reference the same pack. No duplication.

---

Knowledge Atoms

Each atom is a structured unit extracted from a tweet:

{
  "id": "naval_0042",
  "knowledge": "You're not going to get rich renting out your time",
  "original": "You're not going to get rich renting out your time...",
  "url": "https://x.com/naval/status/...",
  "date": "2018-05-31",
  "topics": ["Wealth Creation", "Leverage & Scale"],
  "skills": ["naval-wealth", "naval-leverage"],
  "type": "principle",
  "confidence": "high"
}

Standalone Usage

• System prompt: Paste any reference file into your AI’s system prompt
• RAG: Import atoms.jsonl into your vector database (2,347 entries with topic tags)
• Reference: Read the knowledge packs as standalone methodology docs

---

Difference from dbskill

---

Data Source

• Twitter: @naval
• Tweets: 2,347 original (no replies/retweets)
• Time range: 2010-12-08 to 2026-03-25