VS Code AI Coding

Overview

VS Code has become a hub for AI-assisted coding. This map covers the ecosystem: from GitHub Copilot's core features to advanced techniques like agent skills and context engineering.

Context Engineering

Understanding how to feed the right information to AI coding assistants.

• context-engineering-guide-vscode - Microsoft's official guide to custom instructions, planning agents, and structured workflows
• prompt-files-in-vscode - Reusable, on-demand prompt workflows with .prompt.md files for task-specific AI interactions
• beast-mode-3-1-vs-code-chat-agent - A custom chat mode that fixes GPT-4.1's speed-over-thoroughness tendency using todo lists and sequential thinking

Agents

Understanding VS Code's agent architecture and capabilities.

• using-agents-in-vscode - Four agent types (local, background, cloud, third-party) with unified session management
• mastering-subagents-in-vs-code-copilot - Using #runSubagent for isolated, autonomous task execution in Git worktrees
• introducing-agent-skills-in-vs-code - The Agent Skills open standard: portable workflows that load on demand
• nuxt-skills - Practical implementation of skills for Nuxt development

Development Environments

Isolated, reproducible environments for safe AI agent execution.

• run-your-ai-coding-agent-in-containers - Dev containers for sandboxed AI agent execution with Docker and Codespaces

Building & Extending

SDKs and tools for building your own AI coding experiences.

• copilot-sdk - GitHub's official SDK for embedding Copilot into applications via JSON-RPC
• build-and-deploy-a-cursor-clone - Building a Cursor-like editor from scratch with Next.js

Connections

• context-engineering-guide - Broader context engineering principles beyond VS Code
• writing-a-good-claude-md - Same concepts applied to Claude Code instruction files