Skip to content

youdotcom-oss/agent-skills

Repository files navigation

You.com Agent Skills

Agent skills for integrating You.com's AI-powered search, content extraction, and web capabilities with popular AI agent frameworks and bash-based agents.

These skills provide interactive workflows that guide your AI agent through setting up You.com integrations for SDKs, frameworks, and CLI tools.

Available Skills

ydc-ai-sdk-integration

Integrate Vercel AI SDK applications with You.com tools for real-time web search, AI-powered answers, and content extraction.

Use when:

  • Building AI SDK applications with generateText() or streamText()
  • Adding web search capabilities to your AI agents
  • Extracting and processing web content programmatically

Features:

  • Interactive setup workflow for existing or new projects
  • Three powerful tools: youSearch, youContents
  • Smart integration with existing AI SDK code
  • Support for multiple AI providers (Anthropic, OpenAI, Google, etc.)

ydc-claude-agent-sdk-integration

Connect Claude Agent SDK (Python and TypeScript) to You.com's HTTP MCP server for web search and content extraction.

Use when:

  • Building Claude-powered agents in Python or TypeScript
  • Integrating MCP tools with Claude Agent SDK v1 or v2
  • Adding You.com capabilities to existing Claude applications

Features:

  • Complete templates for Python and TypeScript (v1 & v2)
  • HTTP MCP server configuration patterns
  • Bearer token authentication setup
  • Error handling and validation examples

ydc-openai-agent-sdk-integration

Add You.com MCP tools to OpenAI Agents SDK using Hosted MCP or Streamable HTTP modes.

Use when:

  • Building OpenAI-powered agents with MCP integration
  • Using Python or TypeScript OpenAI Agents SDK
  • Choosing between Hosted MCP and Streamable HTTP deployment

Features:

  • Dual-mode templates (Hosted MCP + Streamable HTTP)
  • Python and TypeScript implementations
  • Complete configuration examples for both modes
  • Tool approval and validation patterns

ydc-crewai-mcp-integration

Integrate You.com's remote MCP server with crewAI agents for web search, AI-powered answers, and content extraction.

Use when:

  • Building crewAI agents that need real-time web access
  • Integrating You.com MCP via MCPServerHTTP or MCPServerAdapter
  • Adding web search and content extraction to existing crewAI workflows

Features:

  • DSL and MCPServerAdapter integration patterns
  • Python implementation with uv/pip setup
  • Bearer token authentication for the remote MCP server
  • Complete crewAI crew and task configuration examples

ydc-langchain-integration

Integrate LangChain applications (TypeScript and Python) with You.com tools for web search, content extraction, and retrieval.

Use when:

  • Building LangChain.js agents with createAgent and initChatModel (TypeScript)
  • Using YouRetriever, YouSearchTool, or YouContentsTool with LangChain (Python)
  • Adding web search or content extraction to existing LangChain workflows

Features:

  • TypeScript: youSearch and youContents via @youdotcom-oss/langchain, structured output with Zod
  • Python: YouRetriever for RAG chains, YouSearchTool + YouContentsTool for agents via langchain-youdotcom
  • Prompt injection defense guidance (W011 trust boundary)
  • Direct invocation and agent-based usage patterns for both languages

teams-anthropic-integration

Use @youdotcom-oss/teams-anthropic to add Anthropic Claude models (Opus, Sonnet, Haiku) to Microsoft Teams.ai applications. Optionally integrate You.com MCP server for web search and content extraction.

Use when:

  • Building Teams.ai apps with Claude models
  • Need streaming, function calling, or conversation memory
  • Optionally want web search capabilities via You.com MCP

Features:

  • Two paths: Basic setup (Claude only) or with You.com MCP
  • Complete templates for new and existing apps
  • Streaming responses and function calling
  • Conversation memory with Teams.ai Memory API

youdotcom-api

Integrate You.com APIs (Research, Search, Contents) into any language using direct HTTP calls — no SDK required.

Use when:

  • Calling You.com APIs directly without an SDK wrapper
  • Need synthesized, cited answers via the Research API
  • Building custom search pipelines with raw Search + Contents data
  • Working in a language without a dedicated You.com SDK

Features:

  • Research API: one call for multi-step reasoning with cited Markdown answers
  • Search API: raw web and news results with filtering, pagination, and livecrawl
  • Contents API: full page extraction (HTML, Markdown, metadata) from any URL
  • Language-agnostic — works with any HTTP client (fetch, requests, httpx, curl)
  • TypeScript and Python reference implementations included
  • JSON Schemas for all request/response shapes

youdotcom-cli

Web search, research with citations, and content extraction for bash agents using curl and You.com's REST API.

Use when:

  • Working with bash-capable AI agents (Claude Code, Cursor, Codex, etc.)
  • Need fast web search with verifiable citations
  • Want simultaneous search + content extraction (livecrawl)
  • Building bash agent workflows with curl and jq

Features:

  • Search works without an API key (free tier)
  • Livecrawl: search + extract content in one API call
  • Research with citations at multiple effort levels
  • Compatible with any bash-based agent

Installation

For Agent Skills Spec Compatible Agents

Install All Skills (recommended):

# Using npm
npx skills add youdotcom-oss/agent-skills

# Using Bun (recommended)
bunx skills add youdotcom-oss/agent-skills

This installs all 8 skills at once:

  • ydc-ai-sdk-integration
  • ydc-claude-agent-sdk-integration
  • ydc-openai-agent-sdk-integration
  • ydc-crewai-mcp-integration
  • ydc-langchain-integration
  • teams-anthropic-integration
  • youdotcom-api
  • youdotcom-cli

Install Individual Skills:

# Install just one skill
npx skills add youdotcom-oss/agent-skills --skill youdotcom-cli
bunx skills add youdotcom-oss/agent-skills --skill ydc-ai-sdk-integration

# Install multiple specific skills
npx skills add youdotcom-oss/agent-skills --skill youdotcom-cli --skill ydc-ai-sdk-integration

Quick Start

Before using any skill, you'll need a You.com API key:

  1. Get API Key: Visit you.com/platform/api-keys
  2. Set Environment Variable:
    export YDC_API_KEY="your-api-key-here"
  3. Request Integration: Ask your AI agent to integrate You.com (see Usage examples below)

Usage

Once installed, your AI coding agent will automatically activate the relevant skill when you request integration. For example:

  • "Integrate Vercel AI SDK with You.com tools"
  • "Set up Claude Agent SDK with You.com MCP"
  • "Add You.com to my Teams app with Anthropic"
  • "Configure OpenAI Agents SDK with You.com MCP"
  • "Integrate You.com MCP with my crewAI agents"
  • "Add You.com tools to my LangChain.js agent"
  • "Integrate You.com Research API into my Python app"
  • "Call You.com Search and Contents APIs directly with fetch"
  • "Add You.com CLI tools to my bash agent"

Each skill provides step-by-step instructions, code templates, and validation checklists.


Skill Structure

Each skill follows the agent-skills-spec format:

skills/{skill-name}/
├── SKILL.md          # Complete workflow with YAML frontmatter
└── assets/           # Code templates (optional, mostly inlined)

Skills are self-contained:

  • YAML frontmatter defines skill metadata (name, description, category, keywords, compatibility)
  • Markdown body contains complete workflow, inline code examples, templates, validation, and troubleshooting
  • Assets directory (optional) for additional templates - most examples are now inlined for immediate visibility

Prerequisites

API Keys:

  • You.com API key: Get yours
  • Provider API keys (Anthropic, OpenAI, etc.) depending on the skill

Documentation

Each skill includes:

  • Prerequisites - Required packages and environment variables
  • Complete templates - Ready-to-run code for Python and TypeScript
  • Configuration examples - Side-by-side comparisons for different modes
  • Validation checklists - Ensure your integration works correctly
  • Troubleshooting - Common issues and solutions

Development

Environment Setup

Create a .env file in the project root with the following API keys:

# Required for all skills
YDC_API_KEY=your-you-com-api-key

# Required for Claude Agent SDK skill
ANTHROPIC_API_KEY=your-anthropic-api-key

# Required for OpenAI Agent SDK skill
OPENAI_API_KEY=your-openai-api-key

Get API keys from:

Skill Evals

Skills are validated by running Claude Code against prompts and checking that the generated integration code passes real API tests.

Run all skill evals:

bun run eval

Run a single skill eval:

bun run eval --skill ydc-crewai-mcp-integration

Run with parallelism:

bun run eval -j 4

Regenerate data/RESULTS.md from existing results (no re-run):

bun run eval:summary

Note: Evals use real API keys from .env and invoke Claude Code as a subprocess to generate integration code. Valid API keys are required.

Eval Structure

data/
├── prompts/
│   └── prompts.jsonl       # One entry per skill variant (id, prompt, grader config)
├── results/
│   └── results.jsonl       # Grader output per eval run (gitignored)
└── RESULTS.md              # Human-readable summary (committed on weekly CI run)

tests/{skill-id}/           # Generated integration code lives here (gitignored)
├── agent.ts                # Example: TypeScript integration file
└── agent.spec.ts           # Tests that validate the generated code

scripts/
├── run.ts                  # Eval orchestrator (clean → harness → grade → summarize)
└── grader.ts               # Scoring logic for generated integration code

Eval IDs and test directories use language suffixes where needed:

  • ydc-claude-agent-sdk-integration-pythontests/ydc-claude-agent-sdk-integration-python/
  • ydc-claude-agent-sdk-integration-typescripttests/ydc-claude-agent-sdk-integration-typescript/
  • ydc-openai-agent-sdk-integration-pythontests/ydc-openai-agent-sdk-integration-python/
  • ydc-openai-agent-sdk-integration-typescripttests/ydc-openai-agent-sdk-integration-typescript/
  • youdotcom-api-pythontests/youdotcom-api-python/
  • youdotcom-api-typescripttests/youdotcom-api-typescript/
  • youdotcom-clitests/youdotcom-cli/
  • Single-variant skills (e.g., ydc-crewai-mcp-integration) use a single test directory

Workflow:

  1. data/prompts/prompts.jsonl contains prompts that trigger each skill
  2. The eval harness runs Claude Code against each prompt, generating code into tests/{skill-id}/
  3. The grader validates the generated code against the test files
  4. Results are written to data/results/results.jsonl and summarized in data/RESULTS.md

CI

Evals run automatically on:

  • Pull requests that change skills/*/SKILL.md, assets, or eval scripts
  • Pushes to main for the same paths
  • Weekly schedule (Monday 06:00 UTC) — results committed back to data/RESULTS.md

Current eval results: see data/RESULTS.md

Linting & Formatting

Check for issues:

# Check all files (TypeScript + Python)
bun run check

# Check only Python files
bun run check:py

Auto-fix issues:

# Fix all files (TypeScript + Python)
bun run check:write

# Fix only Python files
bun run check:write-py

Tools:

  • TypeScript: Biome for linting and formatting
  • Python: Ruff for linting and formatting
  • Python package management: uv

Prerequisites

  • Bun >= 1.2.21 (for TypeScript evals and orchestration)
  • Python >= 3.12 (for Python skill evals)
  • uv (automatically used by Bun scripts for Python)

Contributing

Contributions are welcome! To add a new skill:

  1. Fork this repository
  2. Create a new skill directory in skills/
  3. Add SKILL.md following agent-skills-spec format
  4. Add optional assets in assets/ subdirectory
  5. Add a prompt entry to data/prompts/prompts.jsonl and reference test files in tests/
  6. Test your skill with npx skills add <your-fork>
  7. Submit a pull request

Skill naming convention:

  • Directory name must match name field in YAML frontmatter
  • Use kebab-case (e.g., ydc-ai-sdk-integration)

License

MIT - See LICENSE file for details


Support