Grok search
Skill kaichen/agent-skillset/plugins/grok-search/skills/grok-search
My Agent Skills
npx -y skills add kaichen/agent-skillset --skill grok-searchAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use Grok AI (grok.com) to search and research topics via browser automation. Triggers on "use grok to search", "ask grok", "grok research", "search with grok", or when user wants real-time X/Twitter data analysis, trending topics, or AI-powered web research.
SKILL.md
4.2 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it
Grok Search
Search and research using Grok AI via browser automation with Claude in Chrome MCP.
Prerequisites
- Claude in Chrome MCP extension installed and configured
- Logged into grok.com in Chrome
Quick Start
Direct URL Query (Recommended)
Use URL parameters to submit queries directly without manual input:
https://grok.com/?q={your_query}
This automatically submits the query and starts generating results.
Example Workflow
// 1. Create new tab
mcp__claude-in-chrome__tabs_create_mcp()
// 2. Navigate with query parameter
mcp__claude-in-chrome__navigate({
url: "https://grok.com/?q=your search query here",
tabId: <tab_id>
})
// 3. Wait for results (typically 5-30 seconds for regular, 1-3 minutes for DeepSearch)
mcp__claude-in-chrome__computer({ action: "wait", duration: 10, tabId: <tab_id> })
// 4. Get page text content
mcp__claude-in-chrome__get_page_text({ tabId: <tab_id> })
// 5. Find and click copy button (aria-label="Copy") or (aria-label="复制")
mcp__claude-in-chrome__find({ query: "复制 button", tabId: <tab_id> })
mcp__claude-in-chrome__computer({ action: "left_click", ref: "ref_xxx", tabId: <tab_id> })
// 6. Save to file via clipboard
Bash: pbpaste > result.md
Search Modes
Default Mode (Fast)
- URL:
https://grok.com/?q={query} - Response time: 5-15 seconds
- Best for: Quick questions, simple lookups
DeepSearch Mode (Thorough)
- Requires clicking DeepSearch button before submitting
- Response time: 1-3 minutes
- Best for: Research, comprehensive analysis, multi-source information
To enable DeepSearch:
// After navigating to grok.com (without query param)
mcp__claude-in-chrome__find({ query: "DeepSearch button", tabId: <tab_id> })
mcp__claude-in-chrome__computer({ action: "left_click", ref: "ref_xxx", tabId: <tab_id> })
// Then input query and submit
Common Use Cases
1. Real-time X/Twitter Analysis
Query: "24h内中文热门推特在讨论什么"
Query: "What are trending topics on X about AI in the last 24 hours"
Query: "24h内我关注的中文推特时间线里有哪些热门讨论"
2. News and Current Events
Query: "最近24小时内 Claude 相关的信息有哪些"
Query: "Latest news about [topic] in the past 24 hours"
3. Research Topics
Query: "[Topic] 的最新进展和讨论"
Query: "Comprehensive overview of [topic]"
Handling Results
Method 1: Copy Button (Preferred)
// Find copy button using aria-label
mcp__claude-in-chrome__find({ query: "复制 button", tabId: <tab_id> })
// Click to copy content to clipboard
mcp__claude-in-chrome__computer({ action: "left_click", ref: "ref_xxx", tabId: <tab_id> })
// Save from clipboard
Bash: pbpaste > result.md
Method 2: Direct Text Extraction (Fallback)
// If copy button fails, extract text directly
const text = mcp__claude-in-chrome__get_page_text({ tabId: <tab_id> })
// Then save content via Write tool
Tips
- Check Generation Status: Look for timing indicators like "思考了 Xs" or "Searching" in page text
- Wait Adequately: DeepSearch can take 1-3 minutes; check periodically with
get_page_text - Tab Management: Create new tabs for parallel searches
- Error Recovery: If extension conflicts occur, try
get_page_textas fallback for content extraction - URL Encoding: Complex queries with special characters should be URL-encoded
Troubleshooting
Extension Conflicts
If you encounter "Cannot access a chrome-extension:// URL" errors:
- Try navigating to a fresh URL
- Use
get_page_textinstead of screenshot - Create a new tab and retry
Results Not Loading
- Increase wait time
- Check if Grok is still generating (look for loading indicators)
- Verify you're logged into grok.com
Copy Button Not Working
- Use
get_page_textto extract content directly - Format and save the text content manually
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.