web-search
npx machina-cli add skill valkryhx/google_adk_agent/web-search --openclawWeb Search
Overview
This skill provides capabilities to search the web and extract content from specific URLs. It uses the Tavily API to ensure high-quality, LLM-ready results.
Workflow
- Search: Use when you need to find information, answer questions, or discover resources.
- Extract: Use when you have a specific URL (e.g., from a search result or user input) and need to read its full content.
Usage
1. Search the Web
To search for information, you must have the bash skill loaded and use the bash tool to run the search.py script.
python skills/web-search/scripts/search.py --query "your search query"
Optional Arguments:
--depth: "basic" (default, fast) or "advanced" (comprehensive).--include-answer: If set, returns a generated answer summary.--include-raw-content: If set, returns the full content of the search results.
2. Extract Content from URL
To read the content of a specific URL, run the search.py script with the --url argument.
python skills/web-search/scripts/search.py --url "https://example.com/article"
Examples
User: "What are the latest features in Python 3.13?" Action:
skill_load(skill_id="bash")(if not already loaded)bash(command="python skills/web-search/scripts/search.py --query \"Python 3.13 new features\" --include-answer")
User: "Read this article for me: https://example.com/long-article" Action:
python skills/web-search/scripts/search.py --url "https://example.com/long-article"
Source
git clone https://github.com/valkryhx/google_adk_agent/blob/main/skills/web-search/SKILL.mdView on GitHub Overview
This skill provides capabilities to search the web and extract content from specific URLs using the Tavily API to ensure high-quality, LLM-ready results. It enables real-time information gathering and deep-dives into targeted pages for accurate analysis.
How This Skill Works
You run the search.py script with a query to perform web search results, or you pass a URL with --url to fetch full page content. Tavily API powers the content extraction, returning either summaries or full text based on optional flags such as --depth, --include-answer, and --include-raw-content.
When to Use It
- You need real-time information or the latest news on a topic.
- You want to read a long article or webpage in full without manual browsing.
- You are gathering sources and need links and excerpts for citation.
- You want a quick answer summary alongside web results.
- You are performing a deep-dive into a specific URL or set of URLs.
Quick Start
- Step 1: Load the bash skill (if not already loaded).
- Step 2: Run a search with the query, optionally adding --depth or --include-* flags.
- Step 3: To read a page, run the script with --url and review the results.
Best Practices
- Ensure the bash skill is loaded before using the script.
- Choose depth 'basic' for fast results or 'advanced' for comprehensive extraction.
- Use --include-answer to get a concise summary and --include-raw-content for full text when needed.
- Verify sources by visiting returned URLs and checking publication dates.
- Be mindful of rate limits and terms of service when querying multiple URLs.
Example Use Cases
- User asks: What are the latest features in Python 3.13? Action: load bash (if needed) and run 'python skills/web-search/scripts/search.py --query "Python 3.13 new features" --include-answer'.
- User asks: Read this article: https://example.com/long-article. Action: run 'python skills/web-search/scripts/search.py --url "https://example.com/long-article"'.
- User asks: What AI advancements occurred in 2024? Action: run 'python skills/web-search/scripts/search.py --query "AI advancements 2024" --depth advanced --include-raw-content'.
- User asks: Find sources on climate policy. Action: run 'python skills/web-search/scripts/search.py --query "climate policy sources" --include-raw-content --include-answer'.
- User asks: Compare search results for 'privacy in web scraping'. Action: run 'python skills/web-search/scripts/search.py --query "privacy in web scraping" --include-answer'.
