grok-search_skill

Overview

This skill provides enhanced web search and real-time content retrieval via the Grok API, replacing built-in WebSearch/WebFetch with forced tool routing. It supports both MCP tool invocation and a standalone CLI, enabling up-to-date fact checking, news monitoring, and deep webpage extraction. Use it when you need reliable multi-source aggregation and explicit source citation for time-sensitive information.

How this skill works

The skill routes web queries to GrokSearch tools: web_search returns structured summaries (title, url, description) while web_fetch retrieves full webpage content as Markdown. When MCP tools are available it calls mcp__groksearch__* endpoints; otherwise it falls back to the CLI script. Results are synthesized across sources with required source URLs and date annotations, and the workflow includes automatic retries and parameter tuning.

When to use it

• Real-time news, breaking events, or anything beyond the model cutoff
• Fact checking or multi-source verification across websites
• Extracting full webpage content or parsing a specific URL
• Aggregating technical docs, changelogs, or developer resources
• Replacing default WebSearch/WebFetch behavior with Grok tools

Best practices

• Start with web_search for summaries, then web_fetch key URLs for depth
• Specify platform (e.g., GitHub) and adjust min/max results to focus sources
• Always include source title, URL, and retrieval date for time-sensitive claims
• Retry with broader queries if initial search returns no results
• Run get_config_info when encountering connection or auth issues

Example use cases

• Find the latest announcements and link to original press releases
• Fetch and convert a documentation page to Markdown for analysis
• Aggregate multiple blog posts to verify a technical claim
• Monitor a project repo on GitHub for recent commits or releases
• Switch models or toggle built-in tools during troubleshooting