Workshops

Over the past year, model intelligence has skyrocketed, and agent tooling has improved significantly. However, there's still a gap: most AI agents don't have the procedural, domain-specific expertise needed to do real work inside real products, especially in React apps, where a small behaviour change turns into user-facing regressions.

In this workshop, I'll show you how leveraging Agent Skills changes how you build AI features in React: from prompt tweaking to shipping packaged guides, resources and scripts that turn a general-purpose agent into a product-specific one via the Agent Skills protocol (adopted by OpenAI, Anthropic and Cursor). I'll also show a practical way to do this in React apps without changing your current frameworks or LLM, with evals and versioning in place to prevent regressions.

What you’ll build#

You’ll implement a real React/Next.js AI feature (server route/action + UI) and begin with a large prompt and several tools/MCPs. We will use the Vercel AI SDK for a quick start.

Next step, we will refactor it into an Agent Skills-driven design. You’ll intentionally break the feature in realistic ways (output drift, tool misuse, injection attempts) and then harden it using a small Skills pack. Finally, you’ll add versioning + lightweight evals so updates don’t regress your UI.

By the end, you'd have learned not just the basics of building Agentic AI apps, but taken it a step further by making them reliable with Agent Skills.

React teams are on the front line: agent failures become UI failures. This workshop is designed around React realities: typed contracts, predictable component states, safe fallbacks, and preventing “it worked yesterday” incidents, without requiring you to switch frameworks or lock into one LLM/provider.

Key outcomes#

• A working React/Next.js reference project using Agent Skills
• A reusable Skills pack template (think SOPs + scripts + guardrails)
• A working evaluation framework to catch regressions
• A checklist for shipping agent-powered UI features safely

Workshop prerequisites#

• Intermediate React + basic Next.js familiarity (routes/actions helpful)
• Node.js installed; ability to run a repo locally
• No prior “agent framework” knowledge required

Autonomous coding agents, specifically LLM's wrapped inside what we called harnesses have the possibility to multiply software development efficiency. This efficiency starts from individual usage but eventually changes teamwork dynamics.

As creating code takes less time, other avenues of software development gain significance. This training is about maximising the value of AI-enabled code creation and the resulting changes in teamwork.

What will you learn?#

In this workshop, you'll learn how to configure, run and manage Claude Code. You will also get a sense what other similar coding agents and harnesses can do and how they differ.

We go from running agents individually to creating proper engineering workflows that add value to your whole team, not just software developers.

The main studies are:

• basic setup and usage
• memory and context management
• management of several parallel AI's
• integrate project management and MCP servers to your workflow
• sub-agents, commands, skills
• advance usage patterns
• develop a "compounding engineering" guardrails
• develop a mindset of curiosity and agentic thinking
• pricing and economics of AI agents & tools

After this training you'll be able think ahead of new bottlenecks and how to solve them in novel ways. The goals is not only to enhance all-around development efficiency & team productivity, but to create AI integration momentum for your team!

AI tools#

We use a selection of tools available to anyone. We'll focus more on principles than the details of these tools, so you can carry the knowledge as the tooling eventually changes.

The tools we use:

• Claude Code for agentic development (plus we'll look at Codex, Gemini/Antigravity and Opencode)
• Cursor/VSCode integrations
• Tools like Warp or Conductor to help manage multiple parallel agents
• Linear and Github for task/project management

We will start off with a little theory, then spend most of our time working with Claude Code together.

Who is the workshop for?#

This is for tech-minded people who are not afraid of console and coding tools. For software developers and people working with them. Programming proficiency is a benefit but not a must.

• Software developers
• Designers
• Project managers
• Product Owners

We will be mostly typing in English instead of any programming language.