Pex Agent Skills

Welcome to the official Agent Skills repository for Pex, a Vobile company.

Pex is the global leader in content identification and rights management. This repository provides a collection of modular Agent Skills—standardized instructions and resources that allow AI agents (like Claude Code, GitHub Copilot, or custom LLM workflows) to effectively interface with Pex’s advanced ACR (Automatic Content Recognition) and AI-detection technologies.

📂 Repository Structure

• skills/: The core skill definitions. Each subfolder contains a SKILL.md file that teaches an AI agent how to use specific Pex capabilities.
  • ai-song-detector/: Use API endpoint to identify if an audio file (song) was generated by AI.

🛠 Getting Started

To use these skills, point your agent to the relevant directory or follow the installation instructions within each individual SKILL.md file or install skills using

npx skills add Pexeso/ai-agents

🚀 Why Use These Skills?

By importing these skills into your AI agent's environment, you enable it to:

1. Navigate Pex Tools: Understand Pex’s query syntax and metadata structures.
2. Automate Compliance: Build workflows that automatically flag AI-generated content or copyright issues.
3. Enhance Attribution: Leverage Pex's industry-leading registry to accurately attribute content to rightsholders.

Pex MCP - Model Context Protocol

The Pex MCP server is available at https://mcp.pex.com/mcp. See mcp.json for details.

It provides the following tools:

• detect_ai_song_from_url

  • Detects whether audio from a URL is AI-generated.
  • Parameters: url (string): The URL of the audio file to analyze.
  • Returns: A dictionary containing the detection result, including fields like is_ai and ai_score.

• detect_ai_song_from_base64

  • Detects whether base64-encoded audio is AI-generated.
  • Parameters: file_contents (string): The base64-encoded audio data.
  • Returns: A dictionary containing the detection result.

Client ID and Client Secret must be filled-in when adding the MCP connector to the AI agent. While adding the connector, the AI Agent will show the authorization consent page.

Authorization for the MCP server was tested with:

• Claude
• Grok - choose "Token Auth Method" = client_secret_post or client_secret_basic
• Gemini CLI
• ChatGPT