scrapling-web-extraction

Use Scrapling for lawful, user-directed web scraping, bulk page extraction, AI-targeted page cleanup, screenshots, and dynamic-page fetching.

SKILL.md

GitHub

Scrapling Web Extraction

Use this skill when a task requires extracting content from websites more systematically than ordinary search or manual browser inspection.

适用场景

  • The user asks to scrape or extract data from one or more webpages.
  • A page needs JavaScript rendering before content is available.
  • The task needs batch URL extraction.
  • The output should be Markdown, plain text, HTML, or structured page content.
  • The user wants AI-friendly cleanup of page content.

不适用场景

  • Simple fact lookup where ordinary web search is enough.
  • Local frontend testing where Playwright or the Codex in-app Browser is more direct.
  • Bypassing login, paywalls, CAPTCHA, rate limits, robots rules, or access controls.
  • Collecting private, sensitive, or personal data without a clear lawful basis.

工作流

  1. Clarify the extraction target.

    • Identify URL list, desired fields, output format, and whether JavaScript rendering is needed.
  2. Choose the lightest fetcher.

    • Use HTTP GET extraction for static pages.
    • Use dynamic fetch for pages that require browser rendering.
    • Use AI-targeted extraction when content will be summarized or fed into an LLM.
  3. Respect boundaries.

    • Avoid aggressive rates.
    • Do not attempt to bypass access controls.
    • Stop and ask if the request touches accounts, private data, or restricted content.
  4. Verify output.

    • Inspect a small sample before scaling.
    • Check that key fields are present.
    • Mention any pages that failed, timed out, or looked incomplete.

Example Commands

Static page to Markdown:

scrapling extract get https://example.com output.md --ai-targeted

Dynamic page to Markdown:

scrapling extract fetch https://example.com output.md --ai-targeted --headless

Run MCP server:

scrapling mcp

Codex MCP Config Example

[mcp_servers.scrapling]
command = "C:\\path\\to\\scrapling.exe"
args = ["mcp"]
startup_timeout_sec = 120

Safety Notes

Use Scrapling only for lawful, user-directed collection. Prefer official APIs when available. For websites with terms, rate limits, or robots restrictions, respect those boundaries.