Welcome to the Advanced AI Workshop

You’ve completed the foundation. You can work with AI — prompt it, give it context, refine stories, build features. Now you’ll learn to engineer your AI workflow.

In the foundation workshop, you built a monolithic AGENTS.md and used a single refinement agent. That works. But it doesn’t scale — the file gets bloated, every agent carries knowledge it doesn’t need, and the workflow depends on you remembering every step.

This workshop teaches you to build a modular, team-shareable AI development system using the same VetHub project.

What you’ll build

Exercise	What you’ll do	Time
00: Reconnect & Verify	Verify VetHub and your AI setup from the foundation	~10 min
01: Extract Knowledge into Skills	Break AGENTS.md into focused, reusable skill modules	~25 min
02: Build Specialist Agents	Create backend and frontend agents that use skills	~30 min
03: Build Your Own Agent	Build an agent for your own tech stack	~30 min
04: Project-Specific Overrides	Create project overrides that enrich global agents	~25 min
05: The Orchestrator Pattern	Understand how the orchestrator coordinates agents	~30 min
06: Build a Command	Create a slash command that chains a workflow	~35 min
07: Full Assembly	Test your complete AI workflow end-to-end	~45 min
08: Transfer to Your World	Adapt everything to your own project and stack	Remaining

Total estimated time: ~4 hours (half-day workshop)

Prerequisites

This workshop builds directly on the foundation:

✅ You completed the AI-Driven Development (Foundation) workshop
✅ VetHub is set up and running (backend + frontend)
✅ OpenCode is installed and connected to an AI provider
✅ You have a working AGENTS.md from the foundation

If any of these aren’t ready, start with Exercise 00 to get back on track.

The architecture you’ll build

~/.config/opencode/              ← Global (personal, all projects)
├── AGENTS.md                    ← Universal rules (lean!)
├── skills/                      ← Reusable knowledge modules
│   ├── spring-conventions/
│   ├── testing-standards/
│   └── refine-command/
└── agents/                      ← Global specialist agents
    ├── spring-backend-agent.md
    └── svelte-frontend-agent.md

your-project/.opencode/          ← Project (team-shared, via git)
├── METADATA.md                  ← Project context for orchestrator
├── skills/                      ← Project-specific knowledge
│   └── vethub-patterns/
└── agents/                      ← Project overrides (same name = override)
    └── spring-backend-agent.md  ← Enriches the global version

Two layers: Global (personal) + Project (team). Three building blocks: Skills, Agents, Commands. One coordinator: The orchestrator.

How to work

Start at exercise 00 and work forward at your own pace
Adapt, don’t copy — the exercises give structure, but fill in YOUR conventions, YOUR patterns
Test as you go — after every change, open a fresh AI session and try it
Reference the foundation — if you need to re-read an exercise from the first workshop, do it
Use the hints — every exercise links to hints when you’re stuck

How to use this site

Exercises — the hands-on tasks (work through these in order)
Hints — step-by-step guidance when you’re stuck (try yourself first!)
Resources — facilitator guide and storyline (for the person running the workshop)

Ready? Head to Exercise 00: Reconnect & Verify.