I kept explaining the same requirements to coding assistants: limit tool retries, manage secrets, preserve conversation state and test agent behaviour.
So I turned those lessons into reusable skills for building Google ADK applications on Google Cloud.
The collection covers guardrails, deployment, memory, evaluation, security and front-end integration. It draws on working examples I’ve run and tested.
Explore the [skills](https://github.com/RuslanKhis/agentic-engineering-skills)
You can watch short video explaining them here: [video](https://youtu.be/eKkx7MkcjLI)
Pick one task, try the relevant skill and tell me what the assistant still gets wrong.
Let's connect onLinkedIn Questions, feedback or ideas about the skills or the book? Feel free to write to me at khissamiyev@proton.me.
source & further reading
dev.to — original article
I counted every "I would pay" on Hacker News for 30 days. There were 350, and none of them was a customer.
Your agent's cost problem isn't the model. It's the steps you never measured.
Your Claude Code hook exits 1. It is not blocking anything.