Tabstack Extractor

Overview

This skill enables structured data extraction from websites using the Tabstack API. It's ideal for web scraping tasks where you need consistent, schema-based data extraction from job boards, news sites, product pages, or any structured content.

Quick Start

1. Install Babashka (if needed)

# Option A: From GitHub (recommended for sharing)
curl -s  | bash

# Option B: From Nix
nix-shell -p babashka

# Option C: From Homebrew
brew install borkdude/brew/babashka

2. Set up API Key

Option A: Environment variable (recommended)

export TABSTACK_API_KEY="your_api_key_here"

Option B: Configuration file

mkdir -p ~/.config/tabstack
echo '{:api-key "your_api_key_here"}' > ~/.config/tabstack/config.edn

Get an API key: Sign up at [Tabstack Console]()

3. Test Connection

bb scripts/tabstack.clj test

4. Extract Markdown (Simple)

bb scripts/tabstack.clj markdown ""

5. Extract JSON (Start Simple)

# Start with simple schema (fast, reliable)
bb scripts/tabstack.clj json "" references/simple_article.json

# Try more complex schemas (may be slower)
bb scripts/tabstack.clj json "" references/news_schema.json

6. Advanced Features

# Extract with retry logic (3 retries, 1s delay)
bb scripts/tabstack.clj json-retry "" references/simple_article.json

# Extract with caching (24-hour cache)
bb scripts/tabstack.clj json-cache "" references/simple_article.json

# Batch extract from URLs file
echo "" > urls.txt
echo "" >> urls.txt
bb scripts/tabstack.clj batch urls.txt references/simple_article.json

Core Capabilities

1. Markdown Extraction

Extract clean, readable markdown from any webpage. Useful for content analysis, summarization, or archiving.

When to use: When you need the textual content of a page without the HTML clutter.

Example use cases:

• Extract article content for summarization


• Archive webpage content


• Analyze blog post content

2. JSON Schema Extraction

Extract structured data using JSON schemas. Define exactly what data you want and get it in a consistent format.

When to use: When scraping job listings, product pages, news articles, or any structured data.

Example use cases:

• Scrape job listings from BuiltIn/LinkedIn


• Extract product details from e-commerce sites


• Gather news articles with consistent metadata

3. Schema Templates

Pre-built schemas for common scraping tasks. See references/ directory for templates.

Available schemas:

• Job listing schema (see references/job_schema.json)


• News article schema


• Product page schema


• Contact information schema

Workflow: Job Scraping Example

Follow this workflow to scrape job listings:

Identify target sites - BuiltIn, LinkedIn, company career pages

Choose or create schema - Use references/job_schema.json or customize

Test extraction - Run a single page to verify schema works

Scale up - Process multiple URLs

Store results - Save to database or file

Example job schema:

{
  "type": "object",
  "properties": {
    "title": {"type": "string"},
    "company": {"type": "string"},
    "location": {"type": "string"},
    "description": {"type": "string"},
    "salary": {"type": "string"},
    "apply_url": {"type": "string"},
    "posted_date": {"type": "string"},
    "requirements": {"type": "array", "items": {"type": "string"}}
  }
}

Integration with Other Skills

Combine with Web Search

Use web_search to find relevant URLs

Use Tabstack to extract structured data from those URLs

Store results in Datalevin (future skill)

Combine with Browser Automation

Use browser tool to navigate complex sites

Extract page URLs

Use Tabstack for structured extraction

Error Handling

Common issues and solutions:

Authentication failed - Check TABSTACK_API_KEY environment variable

Invalid URL - Ensure URL is accessible and correct

Schema mismatch - Adjust schema to match page structure

Rate limiting - Add delays between requests

Resources

scripts/

• tabstack.clj - Main API wrapper in Babashka (recommended, has retry logic, caching, batch processing)
• tabstack_curl.sh - Bash/curl fallback (simple, no dependencies)
• tabstack_api.py - Python API wrapper (requires requests module)

references/

• job_schema.json - Template schema for job listings
• api_reference.md - Tabstack API documentation

Best Practices

Start small - Test with single pages before scaling

Respect robots.txt - Check site scraping policies

Add delays - Avoid overwhelming target sites

Validate schemas - Test schemas on sample pages

Handle errors gracefully - Implement retry logic for failed requests

Teaching Focus: How to Create Schemas

This skill is designed to teach agents how to use Tabstack API effectively. The key is learning to create appropriate JSON schemas for different websites.

Learning Path

Start Simple - Use references/simple_article.json (4 basic fields)

Test Extensively - Try schemas on multiple page types

Iterate - Add fields based on what the page actually contains

Optimize - Remove unnecessary fields for speed

See Schema Creation Guide for detailed instructions and examples.

Common Mistakes to Avoid

• Over-complex schemas - Start with 2-3 fields, not 20
• Missing fields - Don't require fields that don't exist on the page
• No testing - Always test with example.com first, then target sites
• Ignoring timeouts - Complex schemas take longer (45s timeout)

Babashka Advantages

Using Babashka for this skill provides:

Single binary - Easy to share/install (GitHub releases, brew, nix)

Fast startup - No JVM warmup, ~50ms startup time

Built-in HTTP client - No external dependencies

Clojure syntax - Familiar to you (Wes), expressive

Retry logic & caching - Built into the skill

Batch processing - Parallel extraction for multiple URLs

Example User Requests

For this skill to trigger:

• "Scrape job listings from Docker careers page"


• "Extract the main content from this article"


• "Get structured product data from this e-commerce page"


• "Pull all the news articles from this site"


• "Extract contact information from this company page"


• "Batch extract job listings from these 20 URLs"


• "Get cached results for this page (avoid API calls)"