How Alex Works

You (in VS Code)
    │
    └── Copilot Chat ──→ Alex (cognitive architecture)
                              │
                              ├── Skills ──→ Domain knowledge loaded on demand
                              ├── Instructions ──→ Behavioral rules and procedures
                              ├── Agents ──→ Specialized modes (Builder, Researcher, Validator)
                              ├── Muscles ──→ Automation scripts for heavy lifting
                              ├── Prompts ──→ Reusable prompt templates
                              └── Memory ──→ Persistent context across sessions

Alex is powered by GitHub Copilot's agent platform. The architecture is a collection of markdown files and scripts that Copilot loads as context, giving it domain expertise, project awareness, and the ability to act: not just suggest.

The Architecture at a Glance

Alex lives in a .github/ folder in your repository:

.github/
├── copilot-instructions.md   ← Alex's identity, routing, and active context
├── skills/                   ← 120+ domain-specific knowledge modules
│   ├── code-review/
│   ├── testing-strategies/
│   ├── security-review/
│   ├── debugging-patterns/
│   ├── ai-writing-avoidance/
│   └── ...
├── instructions/             ← Behavioral rules (always loaded when relevant)
├── agents/                   ← Specialized modes
│   ├── builder.agent.md
│   ├── researcher.agent.md
│   ├── validator.agent.md
│   └── ...
├── prompts/                  ← Reusable prompt templates
├── muscles/                  ← Node.js automation scripts
└── hooks.json                ← Lifecycle event triggers