SKILL.md
Trafilatura - Web Content Extraction
Extract clean, readable text from web pages. Use this instead of WebFetch when you need the actual article content (not a summary).
Installation
uv tool install trafilaturaBasic Usage
Extract article as markdown:
trafilatura -u "https://example.com/article" --markdownExtract as plain text:
trafilatura -u "https://example.com/article"With metadata (title, author, date):
trafilatura -u "https://example.com/article" --markdown --with-metadataWhen to Use
| Scenario | Tool |
|---|---|
| Read an article/blog post | trafilatura |
| Get verbatim page content | trafilatura |
| Quick summary of a page | WebFetch |
| API docs / technical pages | trafilatura |
| News articles | trafilatura |
| Pages behind auth | Neither (need browser) |
Common Patterns
Read and analyze an article:
trafilatura -u "https://blog.example.com/post" --markdownThen discuss the content with the user.
Research a topic across multiple URLs:
# Run for each URL, compare findings
trafilatura -u "https://site1.com/article" --markdown
trafilatura -u "https://site2.com/article" --markdownExtract academic paper info:
trafilatura -u "https://arxiv.org/abs/2512.12345" --markdown --with-metadataOptions Reference
| Flag | Purpose |
|---|---|
-u URL |
URL to fetch |
--markdown |
Output as markdown (recommended) |
--with-metadata |
Include title, author, date |
--no-comments |
Exclude comment sections |
--no-tables |
Exclude tables |
-o FILE |
Write to file instead of stdout |
Troubleshooting
Empty output: Page may be JS-rendered (trafilatura can't handle SPAs) Timeout: Large pages may need --timeout 60 Encoding issues: Add --encoding utf-8
vs WebFetch
- trafilatura: Gets full article text, better for reading/analysis
- WebFetch: Gets AI summary, better for quick lookups
- yomu skill: Similar to trafilatura but with different extraction engine
Use trafilatura when you need the actual content, not a summary.