There's a clear path from typing prompts to running agents — command, skill, subagent, agent. You don't need a CS degree; you need to climb one rung at a time without skipping.
The first rung is the simplest: a saved instruction you can reuse. Instead of retyping your standup summary or your code-review prompt every time, you save it once and call it by name. It's small, but it's your first real piece of leverage.
Reuse beats remembering. The moment you stop re-typing a good prompt, you've started building a system.
A skill is a command that knows when to fire. Instead of you calling it, it triggers itself on the right kind of task and runs the steps. It's packaged know-how — the procedure plus the judgment of when to use it.
This is the jump from 'I have a saved prompt' to 'the right help shows up on its own.'
command: you call it (/weekly-report) skill: it calls itself (fires on 'report')
A subagent is a worker with its own context window, its own instructions, and its own tools. It handles one job inside a bigger effort — a researcher, a reviewer, a drafter — and reports back. Now you're delegating a whole task, not just running a step.
This is where your role shifts from doing the work to assigning it.
At the top, an agent runs a full loop toward a goal: gather context, take action, verify, repeat — until the goal is met. You set the direction and the guardrails; it does the rest and hands you a result to review.
Climb one rung at a time. Each level only makes sense once the one below it is solid — so don't skip. Build quietly, compound deliberately, and you'll be running agents before you ever felt like a 'technical' person.
A one-page map of the four rungs — command, skill, subagent, agent — with the signal that tells you when you're ready to climb.