What Your Agent Knows: Skills, MCP, and the Agentic Development Loop

AI coding agents are only as useful as the context you give them. For GraphQL developers, that means understanding what Skills are, how MCP works, and how the agent loop actually operates under the hood.

This is an interactive session. Questions are encouraged and answered throughout. We’ll cover the architecture, define the terms, and show real Skills in a live demo.

In this session, we cover:

• What an agent is and how the agent loop works

• What MCP is and where it fits in the architecture

• Two types of Skills: general Skills that encode reusable best practices, and project-specific Skills that live inside your codebase

• Where GraphQL fits into an agentic setup and how your existing API is already an asset

• How to write and structure Skills so your agent has the context it needs to do useful work