get-contents_skill

Overview

This skill retrieves and extracts content from web URLs with AI-powered summarization and structured data extraction. It focuses on token- and cost-efficient web scraping by enforcing live crawling and promoting summary- and schema-based extraction. Use it to fetch fresh page content, produce compact summaries, or return validated JSON schemas for downstream processing.

How this skill works

The tool always performs a live crawl (use --livecrawl-timeout 10000) to fetch up-to-date page content. It supports three output strategies: compact toon output for human reading, JSON + jq for programmatic field extraction, and schema-driven summaries (--summary-schema) for structured data. Use --summary instead of full text and wrap schemas in an object to ensure valid extraction.

When to use it

• Fetch fresh, up-to-date content from a web page or set of pages
• Generate concise AI summaries instead of downloading full HTML text
• Extract structured fields (company info, product details) with a JSON schema
• Crawl multiple related pages and summarize each with low token cost
• Pipe results into automation or analytic pipelines using jq

Best practices

• Always include --livecrawl-timeout 10000 to ensure livecrawl is used
• Prefer --summary over --text to reduce tokens and cost
• When extracting structured data, provide a root object schema ({"type":"object","properties":{...}})
• Choose one output approach and stick to it: toon for reading, JSON+jq for programmatic extraction, or schema+jq for validated structures
• Do not pipe toon output to jq — toon is YAML-like and not valid JSON
• Limit extraction with --text-max-characters, --links, and --image-links to control output size

Example use cases

• Summarize a tech article with --summary and --output-format toon for rapid reading
• Extract company_name, main_product, and target_market via --summary-schema and pipe to jq for ingestion
• Crawl a list of competitor pages and return compact summaries to feed into a research dashboard
• Combine search results with targeted get-contents calls to fetch only the most relevant URLs and minimize cost
• Run multi-URL fetches (comma-separated) to batch extract summaries across sites