web-extract - SKILL.md Agent Skill

name: web-extract description: "Extract clean text content from any URL. Uses trafilatura for high-quality extraction, no API key needed." version: 1.0.0 metadata: echo: tags: [Web, Extract, Scraping, Content, URL]

Web Extract

Extract readable text/markdown from any URL. Uses trafilatura — the best Python content extraction library (handles news, blogs, docs reliably).

Quick Usage

pip install trafilatura httpx

import trafilatura

# Fetch and extract in one step
text = trafilatura.fetch_and_extract("https://example.com/article")
print(text)

# With more options
downloaded = trafilatura.fetch_url("https://example.com/article")
result = trafilatura.extract(downloaded, output_format="markdown", include_links=True)

Helper script

python3 scripts/extract_url.py "https://example.com/article"
python3 scripts/extract_url.py "https://example.com" --format markdown --links

Options

Parameter	Effect
`output_format="markdown"`	Markdown with headers
`include_links=True`	Preserve hyperlinks
`include_images=True`	Include image references
`include_tables=True`	Preserve table structure
`favor_recall=True`	Extract more (less precision)

Fallback: httpx + readability

For pages where trafilatura struggles:

import httpx
from readability import Document

resp = httpx.get(url, follow_redirects=True, timeout=15)
doc = Document(resp.text)
title = doc.title()
content = doc.summary()  # HTML, needs html2text for markdown

JavaScript-heavy sites

For SPAs or JS-rendered content, use playwright (optional):

from playwright.sync_api import sync_playwright

with sync_playwright() as p:
    browser = p.chromium.launch()
    page = browser.new_page()
    page.goto(url, wait_until="networkidle")
    html = page.content()
    browser.close()
# Then pass html to trafilatura.extract()

Rate Limits

Be respectful: add 1-2 second delays between requests to the same domain. Set a User-Agent: trafilatura.fetch_url(url, config=config) with custom config.