web-search

Lightweight web search using DuckDuckGo HTML and curl. No browser dependencies, works on Raspberry Pi.

SKILL.md

GitHub

Web Search Skill

Role

You are a Web Search Assistant. Use this skill when the user wants to search the web and get content from the results.

When to Use

USE THIS SKILL when the user asks you to:

  • Search the web or internet
  • Find information online
  • Look up documentation, tutorials, or guides
  • Research a topic

DO NOT USE for local file search or when the user has the information already.

Installation

The tool is already installed at /home/srose/clawdbot/skills/web-search/

cd /home/srose/clawdbot/skills/web-search
npm install  # Already done

Usage

Basic Search

web-search -q "your search query"

With Content Fetching

web-search -q "TypeScript tutorial" --fetch --max-results 3

Options

Option Description
-q, --query Search query (required)
--max-results Number of results (default: 5)
--fetch Fetch full page content for each result

Examples

You: Search for React TypeScript tutorials
Bot: web-search -q "React TypeScript tutorial"

You: Get detailed content from results
Bot: web-search -q "Rust async programming" --fetch --max-results 3

You: Quick answer lookup
Bot: web-search -q "how to install Docker on Raspberry Pi"

Output Format

Search only:

{
  "query": "your search",
  "results": [
    { "title": "Result Title", "url": "https://...", "snippet": "..." }
  ]
}

With --fetch:

{
  "query": "your search",
  "results": [
    { 
      "title": "Result Title", 
      "url": "https://...",
      "snippet": "...",
      "content": "Full page content (truncated to ~1500 chars)..."
    }
  ]
}

Notes

  • Uses DuckDuckGo HTML (no API key needed)
  • Lightweight: uses native fetch + cheerio parser
  • No headless browser = very low memory footprint
  • Works on Raspberry Pi (arm64) without issues
  • ~1-2 seconds per search
  • 10 second timeout for fetching page content