Total Time: 8-9 hours (self-paced) Modules: 7 modules (0-6) Exercises: 25 hands-on exercises Target Audience: Developers at all experience levels
This workshop teaches ways to customize GitHub Copilot. Each module progressively adds customization, demonstrating how they compound.
| Feature | What It Does | Introduced In |
|---|---|---|
| π Repository Instructions | Team patterns in every response | Module 1 |
| π― Agent Plan Mode | Structured thinking before implementation | Module 2 |
| π Custom Prompts | Reusable, shareable prompt files | Module 3 |
| οΏ½ Agent Skills | Domain-specific knowledge loaded when relevant | Module 4 |
| π MCP Servers | External system access (databases, APIs) | Module 5 |
| π€ Custom Agents | Autonomous task specialists | Module 6 |
By Module 6, you'll have a complete understanding of the core customization features and how they compound.
Six personas guide you through the workshop, each representing real developer archetypes:
| Persona | Experience | Role in Workshop | Key Transformation |
|---|---|---|---|
| Sarah | 15 years | Sets the challenge, validates results | Skeptic β Believer β Scaling Advocate |
| David | 20 years | Architecture, code review | "Will AI replace me?" β "AI amplifies my legacy" |
| Marcus | 5 years | DevOps β Application code | Infrastructure only β Full-stack confident β Knowledge sharer |
| Elena | 8 years | Testing, quality assurance | AI-skeptical QA β AI-assisted QA β Quality at scale |
| Rafael | 10 years | Product, specifications | Requirements β Execution bridge β Enabling all teams |
Full personas: modules/00-orientation/PERSONAS.md
| Module | Title | Time | Exercises | Tier | Key Focus |
|---|---|---|---|---|---|
| 0 | The Challenge β Building FanHub | 60 min | 3 | π | Experience Copilot without customization; feel the frustration that drives transformation |
| 1 | Instructions | 50 min | 5 | π | Create magic file + path-based instructions; transform from confused to context-aware |
| 2 | Agent Plan Mode | 90 min | 3 | π | Master structured thinking; use AI to help you think, not just type |
| 3 | Custom Prompts | 90 min | 3 | π | Build reusable .prompt.md library; solved problems stay solved |
| 4 | Agent Skills | 90 min | 3 | π | Teach specialized capabilities via .github/skills/; domain knowledge as AI-accessible resources |
| 5 | MCP Servers | 45 min | 3 | π | Connect to databases and APIs; transform from "knows things" to "can do things" |
| 6 | Custom Agents | 60 min | 5 | π | Bundle instructions, skills, and tools into role-based .agent.md presets |
Total: 7 modules, 25 exercises, ~8-9 hours
| File | Purpose | Created In |
|---|---|---|
AGENTS.md |
Repository-wide team patterns | Module 1 |
docs/ARCHITECTURE.md |
System design documentation | Module 1 |
docs/AI-PLANNING-WORKFLOWS.md |
Planning templates | Module 2 |
.github/prompts/*.prompt.md |
Reusable prompt files | Module 3 |
.github/instructions/*.instructions.md |
File-scoped instructions with applyTo |
Module 4 |
.github/skills/*/SKILL.md |
Agent Skills with domain knowledge | Module 4 |
.vscode/mcp.json |
MCP server configurations | Module 5 |
.github/agents/*.agent.md |
Custom agent definitions | Module 6 |
| Stage | When | What Changes |
|---|---|---|
| Before | Module 0 | Generic, confused suggestions |
| Foundation | Module 1 | Architecture-aware, pattern-following |
| Strategic | Module 2 | Structured planning before implementation |
| Reusable | Module 3 | Prompt libraries, standardized workflows |
| Specialized | Module 4 | Domain expertise via Agent Skills |
| Connected | Module 5 | External system access via MCP |
| Autonomous | Module 6 | Custom agents bundle everything together |
This workshop is open for contributions. Areas of opportunity:
- Additional prompts for the prompt library
- More custom agents for different domains
- Alternative instruction patterns as examples
- Translations to other languages
See AGENTS.md for content guidelines.