Extractor Skill for Claude Code

Zero LLM tokens. Zero hallucination. 100% Deterministic.

Why Not Just Use AI to Fetch and Convert?

When you ask an LLM to "convert this webpage to markdown":

1. Token consumption x2 — AI fetches the page, then AI rewrites it
2. Content drift — Each AI pass can subtly alter meaning
3. Slow — Multiple API roundtrips
4. Inconsistent — Different results every time

This skill runs scripts directly. No AI processing. No token consumption.

Same result every time. No alterations. Exact original!

Your agent delegates the grunt work to deterministic scripts and gets back to thinking.

Your Claude Code agent will love this. Less busywork, more brainwork. Minimal context pollution.

---

What It Does

Input	→	Output	Use Case
Any webpage URL	→	Markdown + images	Save articles, documentation
GitHub specific folder	→	Raw URL list	Feed to NotebookLM, LLMs
GitHub specific folder	→	Downloaded files	Partial clone (skip the whole repo)

---

Installation

One-liner

curl -sL https://raw.githubusercontent.com/professional-ALFIE/extractor-skill/master/install.sh | bash

Manual

git clone https://github.com/professional-ALFIE/extractor-skill.git
cp -a extractor-skill/.claude/skills/extractor ~/.claude/skills/

Standalone Usage

# After install, scripts are at ~/.claude/skills/extractor/scripts/
./extractor_htmlToMarkdown.ts <URL>
./extractor_githubRawLinks.py <GITHUB_URL>
./extractor_githubRawFiles.py <GITHUB_URL> [OUTPUT_DIR]

---

Scripts

1. extractor_htmlToMarkdown.ts

Convert any webpage to clean Markdown with images.

./extractor_htmlToMarkdown.ts --help
./extractor_htmlToMarkdown.ts https://example.com/article
./extractor_htmlToMarkdown.ts https://example.com/article -o article.md
./extractor_htmlToMarkdown.ts https://example.com/article -s  # stdout only

Flag	Description
-o, --output <file>	Save to specified file
-s, --stdout	Print to terminal (no file/images saved)
-h, --help	Show help
-v, --version	Show version

Output:

Article-Title.md
Article-Title_images/
├── image1.png
└── ...

Under the hood: Readability.js (content extraction) + Turndown (HTML→MD) + highlight.js (code language detection)

Requirements: bun — dependencies auto-install on first run

---

2. extractor_githubRawLinks.py

Get raw URLs for files in a specific GitHub path.

Why not git clone? You only need one folder, not 500MB of repo.

./extractor_githubRawLinks.py https://github.com/owner/repo/tree/main/docs

Output:

https://raw.githubusercontent.com/owner/repo/main/docs/guide.md
https://raw.githubusercontent.com/owner/repo/main/docs/api.md

• Auto-copied to clipboard (macOS)
• Supports any branch, tag, or commit

Requirements: Python 3 + requests

Tip: Set GITHUB_TOKEN to increase rate limit (60 → 5000 req/hour)

---

3. extractor_githubRawFiles.py

Download files from a specific GitHub directory — skip the full clone.

./extractor_githubRawFiles.py https://github.com/owner/repo/tree/main/docs ./local-docs

Auto-ignores: node_modules/, .git/, __pycache__/, dist/, build/, .next/, venv/, .pyc, .exe, .dll, .so

Requirements: Python 3 + requests

Tip: Set GITHUB_TOKEN to increase rate limit (60 → 5000 req/hour)

---

Use with Claude Code

Once installed, just ask:

• "Convert this URL to markdown: https://example.com"
• "Get raw links for https://github.com/owner/repo/tree/main/docs"
• "Download files from https://github.com/owner/repo/tree/main/src"

---

License

MIT