Lessons / Power

Saved skills & slash commands

The gap between a power user and a true operator is the library. Every procedure you run repeatedly should become a verb you can invoke by name. That's the unit of compounding leverage — and it's why the most effective people accumulate dozens of custom commands.

A good skill follows a simple shape: load the relevant state, verify it, then act. Make it idempotent so it's safe to re-run.

One distinction to bake in: freeze the procedure, pull the volatile data fresh. The saved command holds the steps — the shape that doesn't change. It should re-fetch the live inputs (today's calendar, the current inbox, this week's numbers) every time it runs, not bake in yesterday's. A command that caches stale data is worse than no command.

Try it now

Pick a sequence you run often and ship it as a named command:

I do this same sequence often: [describe the steps]. Turn it into a reusable, named command I can invoke by name. Follow the load-state → verify → act shape, and make it idempotent so it's safe to re-run.

Where it lives depends on your AI. In Claude Code this is a literal /slash-command. On claude.ai or the desktop app there's no native slash command yet — save the procedure as a Project instruction (or a saved prompt / memory) and invoke it by a short name like "run my morning brief." Same idea, same compounding leverage; don't get hung up on the slash. The point is a named, reusable capability, not the punctuation.

You've got it when…

You invoked a working custom command by name and it ran the whole procedure end to end. Each one you build makes the next workflow cheaper.

Quiz — did it land?

Your tutor checks these before marking the lesson complete:

1. What three-step shape should a good skill follow?
2. Ship one custom command and invoke it by name — what procedure did it run end to end?