Querying Web Search Agent

Create self-contained query documents that include title, problem summary, current configuration, symptoms, what you've tried, specific questions, and additional context. This ensures the search agent has all necessary information to perform precise research without relying on your conversation history or codebase.

Structure your query with explicit sections: describe the problem clearly, list your current configuration and setup, detail symptoms you're experiencing, enumerate solutions already attempted, ask specific questions, and state what information or resources you need from external sources like documentation or forums.

Use standalone queries when you need expert guidance from external documentation, forums, or official sources but the search agent cannot access your codebase or prior conversation context. This applies to technical troubleshooting, configuration research, and dependency investigation.

Yes. By creating structured, self-contained query documents in markdown format, you provide a separate Claude instance with web search capabilities all the context needed to research your problem independently without access to your local environment or conversation history.

Use markdown-ready files with standardized sections: title, problem summary, current configuration, symptoms, what you've tried, specific questions, and additional context. This tool-agnostic format ensures queries are complete and effective for external research across documentation and community sources.

Organize your problem into discrete sections covering what you're trying to accomplish, your current setup, observed behavior, attempted solutions, and explicit questions needing answers. This structure helps web search agents identify relevant documentation and forum discussions without needing your project context.