Angular Agentic Engineering: From Spec to Ship

Agenda

• Date: 30.10
• Duration: 8 hours
• Workshop Lead: Fanis Prodromou (GDE)

Hands-on throughout, on a real Nx/Angular codebase. All hands-on work is based on Claude Code. If you use another agentic tool like Cursor or Antigravity, you’ll get a translation handout to follow along.

1. 01

   Foundations: Vibe Coding vs. Agentic Engineering, the Lifecycle & Context

   • Vibe Coding vs. Agentic Engineering — Place your current AI habits on the spectrum from casual prompting to disciplined agentic engineering, and learn when each is the right call for the stakes at hand.
   • The SDLC, Reshaped by AI — See how AI transforms every phase of the software development lifecycle, and adopt a shared, AI-driven SDLC as the common vocabulary for the day.
   • Conductor vs. Orchestrator — Understand the developer’s evolving role, and why agentic work is a genuinely different mode from autocomplete-style tools.
   • Context Engineering: Static vs. Dynamic — Move beyond prompt engineering to the harness that makes an agent reliable — context files, Skills, permission guardrails, and MCP servers, and which category each one belongs to.
   • Skills Live, Before & After — Install an official Angular skill mid-session and feed a real design straight into the agent, watching context transform the output on an actual component.
2. 02

   Spec to Feature: From Idea to Working Code

   • Skill Anatomy — Install a production skills plugin and dissect real, battle-tested skills to see exactly how they’re built.
   • Spec-Driven Development — Turn a hardcoded, static section into a real feature by writing the spec first, not the code.
   • Planning & Task Breakdown — Decompose that spec into small, ordered, verifiable slices before building anything.
   • Incremental Build with NgXs — Implement one vertical slice at a time — state, list, and a fully validated form — built against a real design.
3. 03

   Scaling Yourself: Custom Skills, Sub-agents & Ultraplan

   • Encode Your Conventions — Author a companion Skill that teaches the agent your workspace’s own architecture, state shape, and boundary rules.
   • Delegating to Sub-agents — Hand focused work such as code review to specialized sub-agents running alongside your main session.
   • Ultraplan — Plan a genuine multi-library change as a reviewable plan artifact before anything touches the codebase.
   • Boundary-Safe Scaffolding — Wire a feature end to end across libraries and confirm your architectural boundaries held.
4. 04

   The Daily Grind: Debugging, Testing & Review

   • AI-Assisted Code Review — Run a review pass that surfaces dead code, risks, and refactoring opportunities in real code.
   • Refactoring with Confidence — Restructure existing code while keeping its behavior intact, with the agent doing the heavy lifting.
   • Catch What the Agent Misses — A deliberately, subtly-wrong output you have to spot before it ships: the skill that doesn’t rust.
   • Testing with Realistic Data — Write meaningful unit tests covering state logic and dynamic form validation, backed by realistic mock data.
5. 05

   Governance, Guardrails & Wrap-up

   • What Stays Human — Walk back through which parts of the lifecycle stayed human-owned all day, and why that answers the “will my skills rust?” worry.
   • The Governance Checklist — Leave with a consolidated, take-home reference of what to double-check at every stage.
   • Shipping Discipline — Commit hygiene and safe shipping practices, without the CI/CD rabbit hole.
   • Your Tool Rosetta Stone — A handout mapping every technique to its equivalent in other agentic tools like Cursor and Antigravity.

Takeaways

1. 

   Master the AI-Driven SDLC

   See how AI reshapes every phase of the software development lifecycle, from a traditional SDLC to an AI-driven one. Learn which parts of the process stay human-owned and which you can confidently delegate to an agent.

2. 

   Know Where You Sit on the Spectrum

   Move deliberately along the spectrum from casual "vibe coding" to disciplined agentic engineering. Learn to read the stakes of a task and choose the right level of structure, verification, and human judgment for it.

3. 

   Engineer Context, Not Just Prompts

   Understand why clever prompts alone fall apart on a real codebase, and why the shift is from prompt engineering to context engineering. Master the static and dynamic context - the harness - that makes an agent reliable inside your architecture instead of guessing around it.

4. 

   Encode Your Architecture into Skills

   Turn your team's hard-won conventions and boundaries into reusable Skills the agent follows every single time — so output quality stops depending on whether you remembered to re-explain yourself in the prompt.

5. 

   Plan and Ship Real Features

   Build a complete feature the way a senior engineer does — spec first, then plan, then supervised, incremental delivery — using sub-agents and multi-step planning to scope even genuinely complex, cross-boundary changes before any code is written.

6. 

   Keep the Skill That Doesn't Rust

   Practice catching the subtle mistakes agents make in review and testing — the ones that pass silently and slip through. The judgment to know when the AI is wrong is the exact skill this day sharpens.

Your Trainer

Fanis Prodromou

Google Developer Expert

Google Developer Expert (GDE) for Angular and Senior Front-End Developer. He is passionate about sharing his knowledge through his YouTube channel, speaking at conferences and meetups, and technical articles. He also stays current in the ever-evolving tech world by attending conferences and experimenting with new technologies, and is a co-organizer of the Angular Athens meetup in Greece.

Get your tickets

Join us for an online workshop. Secure your spot today!

FAQ