distil_skill

Overview

This skill provides discoverable, consistent access to the Distil proxy so agents can fetch web pages as clean Markdown and perform web searches without manual URL building or header handling. It saves tokens by returning stripped, LLM-friendly Markdown content and offers commands for screenshots, rendering, raw content, and cache control. Use an API key and optional extra headers to customize behavior.

How this skill works

The skill calls the Distil proxy to fetch or search web content and returns results as concise Markdown, including titles, URLs, descriptions, and page body. It supports rendering JavaScript-heavy pages, taking screenshots, bypassing cache, and returning raw page content when needed. Extra headers and a proxy base URL can be configured via environment variables for fine-grained control.

When to use it

• When you need compact, token-efficient page content for LLM consumption.
• When you want search results returned with summaries and Markdown-ready page bodies.
• When scraping or reading documentation, blogs, or technical pages programmatically.
• When pages require client-side rendering before extraction (single-page apps).
• When you need screenshots or want to bypass cached content for freshness.

Best practices

• Set DISTIL_API_KEY in your environment for authenticated access.
• Use DISTIL_EXTRA_HEADERS to control caching, token limits, or other proxy features.
• Prefer distil fetch for readable Markdown and distil raw for exact original content.
• Use distil render for JavaScript-driven pages before extracting Markdown.
• Respect robots.txt and site terms; avoid excessive automated requests to a single host.

Example use cases

• Researching APIs or docs: distil fetch converts long docs into token-efficient Markdown for summarization.
• Competitive research: distil search returns ranked results with Markdown snippets for quick analysis.
• Content archiving: fetch and store clean Markdown snapshots of pages for long-term backups.
• Data extraction from SPAs: distil render then fetch to capture rendered DOM content.
• Automated reporting: take a screenshot of a page to include visual evidence alongside extracted text.