youdotcom-cli_skill

Overview

This skill integrates You.com’s CLI into bash-based AI agents to perform web search with livecrawl (search+extract) and to extract content from URLs using the @youdotcom-oss/api (ydc) tool. It provides a lightweight, scriptable interface for real-time web data and content extraction within agentic workflows. The skill is designed for TypeScript-based agent tooling but works from any Bash environment that can run the ydc CLI.

How this skill works

The skill uses the ydc CLI to issue searches and to request livecrawl extraction when full page content is needed. Commands always run with --json and a --client identifier so agent context and error reporting are preserved. For direct URL extraction it calls ydc contents with a list of urls and desired formats (markdown recommended). Use ydc --schema to discover parameters programmatically and build dynamic queries.

When to use it

• When an agent needs fresh web results or live full-page content in real time
• When extracting structured article or documentation content from URLs
• When building bash-based agent workflows that require a simple web-search CLI
• When you need search + immediate crawl (livecrawl) instead of separate scraping
• When integrating web data into Claude, OpenAI, Vercel AI SDK, or Teams.ai agents

Best practices

• Always set YDC_API_KEY as an environment variable before running commands
• Include --client YourAgentName on every command to aid support and tracing
• Use --schema to enumerate supported fields and construct valid json payloads
• Prefer livecrawl_formats: markdown for clean, readable extracted content
• Handle exit codes (0 success, 1 API error, 2 invalid args) and add retry/backoff for 429

Example use cases

• Agent needs the latest news: ydc search with freshness filter and parse results with jq
• Documentation ingestion: ydc search --json '{"query":"docs","livecrawl":"web","livecrawl_formats":"markdown"}'
• Bulk URL extraction pipeline: ydc contents --json '{"urls":[...],"formats":["markdown"]}' and store markdown
• Adaptive agent tool selection: if user supplies URLs use contents, else use search with livecrawl
• Integrating into CI or serverless functions to fetch and normalize web content for downstream models