Production Emergency: Fix a critical bug and modernize the system using GitHub agents.
Monday, 3:00 PM. FlavorHub's search feature is crashing for 30% of users. 500+ error reports. CTO demands a fix AND a lasting solution.
This workshop walks you through exactly that — using GitHub's agent ecosystem to go from chaos to production-ready.
| Tool | What You Do |
|---|---|
| Agent Skills | Diagnose production errors automatically |
| Custom Agents | Build specialist agents with domain expertise |
| Spec Kit | Drive governance, planning, and code generation from a specification |
| Copilot CLI | Run compliance checks and quality gates |
| GitHub MCP | Integrate agents directly into your repo for continuous improvement |
- Python 3.11+ installed
- Valid GitHub Copilot subscription (Individual, Business, or Enterprise)
- VS Code with GitHub Copilot Chat extension
- Git CLI for version control
- Speckit installed and configured
- Copilot CLI
- GitHub CLI for MCP interactions
- Node.js 18+ (for Copilot CLI)
- GitHub MCP
- Python 3.11+ (for running the recipe-manager app)
- UV (for installing Spec Kit)
Starter Code Provided: FlavorHub Recipe Manager brownfield application with intentional issues for learning
The workshop is designed with flexible learning paths:
Complete this path to experience the full Spec Kit workflow — from bug reproduction to a working fix:
| Step | Exercise | Topic | Time |
|---|---|---|---|
| 1 | Exercise 1 — Part 1 | Reproduce the Bug | ~2 min |
| 2 | Exercise 2 | Spec Kit Setup | ~5 min |
| 3 | Exercise 3 | Constitution & Specification | ~8 min |
| 4 | Exercise 4 | Plan & Tasks | ~7 min |
| 5 | Exercise 5 — Part 1 | Implement Validation Module | ~7 min |
| — | Total | Everyone should complete this | ~29 min |
Explore these exercises in any order. They are fully self-contained with no dependencies on each other:
| Exercise | Topic | Time | Focus |
|---|---|---|---|
| Exercise 1 — Part 2 | Issue Analyzer Skill | ~3 min | Error diagnosis |
| Exercise 5 — Part 2 | Implement Remaining Modules | ~5 min | Complete refactor |
| Exercise 6 | Validation & Quality Gates + Code Review | ~15 min | Pre-deployment checks |
| Exercise 7 | Custom Agent — Search Architect | ~8 min | Domain expertise |
| Exercise 8 | Agent Skills — GitHub MCP | ~7 min | Issue automation |
| Exercise 9 | Copilot CLI — Custom Agent Creation & Delegation | ~8 min | Alternative custom agent creation |
Pro Tip: Complete the mandatory core path first (~29 min), then pick optional exercises based on what interests you most. If time is short, you can revisit them after the workshop.
The workshop has been tested with the following AI models on GitHub Copilot:
Claude Sonnet 4.6,GPT-5.3-codex. Results may vary with different models. If you encounter issues, try switching to one of these models in your Copilot settings.
Once your repo is created and cloned, start here: Exercise 1 — Reproduce the Bug