web-search

Retrieve grounded web search results with cited answers via LiteLLM endpoints.

2|1|Updated Apr 9, 2026
One-click install
npx skills add https://github.com/timgranlundmarsden/claude-agent-flow --skill web-search-timgranlundmarsden
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: web-search
Source: https://github.com/timgranlundmarsden/claude-agent-flow/tree/main/skills/web-search
Command: npx skills add https://github.com/timgranlundmarsden/claude-agent-flow --skill web-search-timgranlundmarsden

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the problem of needing fresh, grounded web research when a model's internal knowledge is stale or insufficient, while preserving citations for verification.

Core Features & Use Cases

  • Live Search Access: Queries a grounded-search endpoint to retrieve current information from the web.
  • Cited Answers: Returns a synthesized response with source links so findings can be checked quickly.
  • Flexible Output Modes: Supports answer, search-only, and raw JSON output for different research and debugging needs.
  • Use Case: A developer can ask for the latest library behavior, compare vendor docs, or gather current background research without leaving the agent workflow.

Quick Start

Ask the agent to use the web-search skill to find the latest information on your topic and return a cited answer.

Frequently Asked Questions about web-search

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get live web search results with citations inside Claude Code?

You can retrieve live web search results with source citations by querying a grounded-search endpoint. This approach returns a synthesized response containing current information and source links for verification when internal model knowledge is stale.

Do I need LiteLLM to enable grounded search for technical documentation lookup?

Yes, you need a LiteLLM-compatible chat/completions endpoint with grounded-search support. You must also configure your model and API key settings to query the endpoint successfully for current technical documentation.

What's the best way to fetch current library behavior without leaving an agent workflow?

Use a grounded search request to fetch current library behavior directly within your agent workflow. This method queries the web for up-to-date vendor docs and synthesizes a cited answer without interrupting your development process.

Can I get raw JSON output from web search results instead of a synthesized answer?

Yes, flexible output modes support search-only and raw JSON formats alongside synthesized answers. These modes accommodate different research needs and allow debugging of the raw response data returned by the grounded-search endpoint.

What dependencies are required to assemble and parse grounded search requests with curl?

You need jq and curl installed to assemble requests and parse responses. These command-line tools handle the request construction and JSON response parsing required to interact with the LiteLLM-compatible endpoint.

When should I use a web search skill instead of relying on internal model knowledge?

Use grounded web search when internal model knowledge is insufficient or stale. It is specifically designed for live research, citation gathering, and current-information tasks like technical investigation or comparing vendor documentation.