Agentic Engineering with Svelte

Syllabus

1. Introduction to Agentic Engineering

   We'll introduce agentic engineering and the current challenges of using AI effectively with Svelte applications.

2. Project setup

   We'll create a SvelteKit application with a recommended development stack, including TypeScript, ESLint, Prettier, testing, authentication, database integration, and deployment.

3. Svelte MCP server

   We'll explain what MCP is, how it improves AI-generated code, and how to use MCP tools for documentation lookup, code validation, auto-fixing, and playground generation.

4. AI coding best practices

   We'll cover when to use AI, how to manage LLM context efficiently, and how skills, MCP servers, and context optimization fit together.

5. Working with AI agents

   We'll use sub-agents for parallel and isolated tasks, configure agent instructions in AGENTS.md, and discuss security considerations for MCP servers and AI skills.

6. AI-assisted UI/UX workflow

   We'll explore how to convert a Figma design into code and how to use agents without losing design intent or implementation quality.

7. Development workflow

   We'll establish a Git workflow for reviewing AI-generated changes, practice Test-Driven Development with AI agents, build reusable components following TDD principles, and explain why extensive testing and validation are essential for both engineers and agents.

8. Prompting strategies

   We'll compare a direct prompt with a structured planning approach by solving the same complex task both ways, then analyze when planning produces better outcomes than immediately generating code.

9. Kanban board implementation

   We'll kick off a Kanban board implementation using the established AI-driven workflow. Along the way, we will update AGENTS.md when necessary, discuss what belongs there vs. more specialized documentation, create codebase-specific skills, and write ad hoc ESLint rules that keep the agent on track.

10. Code review with AI

    We'll explore AI-assisted code review options, set up Greptile, and use it to review and improve generated code.

11. Repository automations

    We'll set up repository automations that automatically triage issues and improve CI workflows.

12. Observability and debugging

    We'll integrate Sentry into the application and use it to diagnose, debug, and fix runtime issues in an AI-assisted development workflow.

13. Loop engineering

    We'll end the workshop exploring every manager's dream: giving the agent full autonomy to own the complete development cycle from making a code change through merging the PR and deploying it. We'll set up the necessary tools and infrastructure and uncover where the limits are.