google-search

What This Skill Does

The google-search skill is a powerful, integrated search tool designed to provide OpenClaw agents with high-fidelity information retrieval. It functions as the primary search engine for the platform, leveraging the Gemini 2.5 Flash model equipped with Search Grounding capabilities. Unlike standard web search tools, this skill excels at synthesizing complex information into human-readable answers, complete with accurate, numbered citations. It also supports raw data retrieval through the Custom Search JSON API, allowing for flexibility between synthesized answers, structured raw link lists, and image search queries. It is specifically optimized for Hebrew language support, making it an essential utility for users requiring high-quality results in both Hebrew and English contexts.

Installation

To integrate this skill, ensure you have a valid GOOGLE_API_KEY obtained from the Google Cloud Console, where you must enable both the Gemini API and the Custom Search JSON API. For raw search and image modes, you must also define the GOOGLE_CSE_CX (Custom Search Engine ID). Configuration is handled within the OpenClaw environment settings by defining the mandatory environment variables. The skill can be installed directly via the CLI: clawhub install openclaw/skills/skills/shaharsha/google-search-grounding.

Use Cases

• Grounded Q&A: Perfect for fact-checking, summarizing current news, or answering complex questions where sources are required for verification.
• Research and Link Discovery: Use the raw mode when you need a curated list of URLs and site snippets for deep-dive research or building reference lists.
• Visual Retrieval: Utilize the image mode for fetching visual media, thumbnails, or references related to specific subject matters.
• Multilingual Contexts: Ideal for users working in non-English environments, particularly Hebrew, where standard models may struggle with colloquialisms or local syntax.

Example Prompts

1. "Search for the latest technological advancements in AI for healthcare and provide a summary with sources."
2. "מהן התחזיות הכלכליות לשנת 2025 בישראל?"
3. "Give me a list of 5 high-quality sources for learning advanced Rust programming."

Tips & Limitations

• Preferred Priority: Always default to search mode (grounded) for the best balance of accuracy and convenience; reserve raw mode specifically for instances where URLs are required without model-level interpretation.
• API Costs: Be mindful that frequent use of the Gemini Search Grounding API can incur costs; monitor your Google Cloud usage.
• Configuration: Ensure the GOOGLE_SEARCH_LANG is set correctly in your config to avoid retrieving results in the wrong language. If the API returns no results, check that your GOOGLE_CSE_CX ID is properly scoped to the entire web or your intended search context.