Data Extraction

---
name: cu-data-extraction
description: |
  Extract structured data from websites using Computer Use. Trigger when the
  user needs to scrape, extract, or collect data from web pages. Phrases:
  "extract data", "scrape this page", "get the table from", "pull pricing data".
allowed-tools:
  - computer
---

# Data Extraction (Computer Use)

Extract structured data from any website using visual browser automation.
Handles tables, listings, profiles, pricing pages, and any structured
content that is visible on screen.

## Workflow

1. **Navigate** to the target URL
2. **Identify** the data structure on the page
3. **Extract** data systematically (row by row, card by card)
4. **Structure** output as JSON, CSV, or markdown table
5. **Paginate** if data spans multiple pages

## Output formats

- **JSON** - For API-like structured data
- **CSV** - For spreadsheet-ready tabular data
- **Markdown table** - For documentation

## Example usage

- "Extract the pricing table from competitor.com"
- "Scrape all job listings from this careers page"
- "Pull the product catalog into a CSV"
- "Get all author profiles from this directory"

## Tips

- Start with a screenshot to understand the page layout
- Extract one complete row/card first, then repeat the pattern
- Handle pagination by clicking "Next" and repeating
- Check for dynamic loading (scroll to trigger lazy-loaded content)

## Requirements

- Computer Use enabled in your Claude environment
- Target website must be accessible