TLDR: An Agentic Engineer should have one philosophy in mind. That is to automate everything. Never are you supposed to type code. Never are you supposed to write in the CLI. Your keyboard was never the tool, you were. Now design the intelligence that writes it.

When generating code using AI, there’s possibly one set of really good advice. Automate everything, and be patient. Working with agents you should have them generate all the code for you.

Take up the Agentic Engineer Ethos, “No Human Typed Code!”. Your keyboard was never the tool, you were. Now design the intelligence that writes it.

This also applies to the CLI as well.

As developers, as humans, we just want to create the code. We don’t really care about the specs. We don’t really care about wanting to plan it, even if it’s just a small change. So I urge you to take patience and have the patience that you need, so then you can automate your workflow and everything is performed by the agents.

Think in terms of Prompts and Programs. Every time you need something performed I whip out a prompt and get the Agent ready to make a prompt or skill for me.

Remember the methodology, the philosophy, and the concepts, and 
tasks used in the following discussion that we are about to have, 
and the work that you performed, please think about how you would 
create a prompt and the scripts, to hand to somebody, who has no 
prior knowledge of the system or tools and concepts applied, that 
would instruct them to perform the task that you just performed. 
Ensure that it is generic, so that it can apply to any situation.

This way we create a prompt, script, or skill after performing a task, and it is ready for next time. This is great because I can build it into my automation scaffolds.