firecrawl-scrape_skill

Overview

This skill extracts clean, LLM-optimized Markdown from one or more URLs, including JavaScript-rendered single-page applications. It runs concurrent scrapes, supports selective tag inclusion/exclusion, and can wait for client-side rendering before capturing content. Use it whenever you need readable, structured page content for downstream analysis or indexing.

How this skill works

The skill loads pages (headless browser when needed) and renders JavaScript so dynamic content is captured. It strips navigation, sidebars, and footers when requested and can include multiple output formats (markdown, HTML, links, screenshots). Multiple URLs are fetched concurrently and returned as concise, LLM-friendly Markdown or JSON when multiple formats are selected.

When to use it

• You provide a URL and ask to “scrape”, “fetch”, “grab”, “get the page”, “extract from this URL”, or “read this webpage”
• Page content is needed from a static site or an SPA that requires JS rendering
• You need clean, readable Markdown for summarization, indexing, or ingestion into an LLM pipeline
• You want to scrape multiple pages concurrently and save outputs to files
• You need main content only (no nav/footer) or selective HTML tag extraction

Best practices

• Try scrape before a full browser automation step — it handles most static and JS-rendered pages without interaction
• Quote URLs in shells to avoid issues with ? and & characters
• Use --only-main-content to remove nav, footer, and sidebars for focused text extraction
• Use --wait-for <ms> when pages rely on delayed JS rendering or client-side data loading
• Request multiple formats (markdown,links) to get content and link metadata in a single JSON output

Example use cases

• Extract article text as Markdown for summarization or content analysis
• Pull documentation pages into a knowledge base with consistent Markdown formatting
• Batch-scrape multiple landing pages to build a link index or sitemap snapshot
• Capture SPA-rendered product pages for price monitoring or feature extraction
• Save clean Markdown versions of search results pages for downstream LLM processing