Working Smarter: Skills, Reusable Prompts, and Power Habits

Every Friday, Dana typed the same thing. She managed a twelve-person team, and each week she opened her AI tool and explained the whole situation from scratch: who was on the team, the tone her manager liked, the format their weekly update had to follow, the three projects nobody was allowed to call “delayed.” Six paragraphs of setup before she could ask for the one thing she actually wanted. She did this for four months. Then a coworker watched over her shoulder and asked why she didn’t just save it. Dana saved it once. She has not retyped it since.

Why this matters to you

By now you can write a strong prompt. You know how to give the task, the context, and the format, and you know how to push back on a first draft until it’s right. That’s real skill, and most people never get there.

Here’s the next step: stop starting from scratch. The same tasks come around again and again. The weekly email. The permission slip. The practice plan. The thank-you note to a donor. Every time you re-explain your situation, you’re paying the same setup cost twice, ten times, a hundred times. This chapter is about paying it once.

Two tools do the heavy lifting. A prompt library saves the prompts that worked so you can reuse them. A skill saves your context and preferences so the AI applies them automatically, without you retyping a word. Together they turn AI from something you operate into something that already knows how you work.

Stop repeating yourself: the prompt library

A prompt library is exactly what it sounds like: a place where you keep the prompts that worked so you never have to reinvent them. It isn’t special software. It’s a note, a document, a page in whatever app you already use for lists. The point is that it exists and you add to it.

Keep three things for each entry. First, the task in a few words, so you can find it later (“weekly team update,” “reply to an upset parent,” “turn my messy notes into an agenda”). Second, the exact prompt you used, copied and pasted, punctuation and all. Third, one line on what made it work: the detail that turned a bland answer into a good one. Maybe it was the sentence “keep it under 150 words.” Maybe it was pasting in last week’s version as an example. That one note is what makes the prompt reusable instead of lucky.

The reason to record what made it work is simple. A prompt you can’t explain is a prompt you can’t improve. When you know the moving parts, you can swap the details next week and get the same quality.

Try It Now Open a fresh note and title it “My Prompt Library.” Now look back through your work in this course so far and find the single best prompt you’ve written: the one that gave you an answer you actually used. Paste it in. Above it, write the task in a few words. Below it, write one sentence: what made this prompt work? That’s entry number one. You now have a library. What did you notice about your own best prompt when you had to explain why it worked?

What “skills” are (and why every tool has them)

A prompt library saves the words. A skill saves the setup.

A skill is a set of instructions you write once that your AI tool applies automatically, every time, without you retyping them. Think of it as briefing a new assistant on their first day. You explain who you are, what you’re working on, how you like things done, and the rules that never change. A good assistant remembers all of it. From then on, you just hand them the task and they fill in the rest, because they already know the background.

That’s what a skill does. Instead of opening each session by re-explaining that you coach a middle-school soccer team, that parents get nervous about schedule changes, and that every message should end with the field address, you write that down once. After that, the AI loads it on its own. You ask for “a message about Saturday’s rained-out game,” and the reply already sounds like you, already reassures the nervous parents, already includes the field address for the makeup.

Nearly every AI tool has a version of this. The names differ, the button is in a different place, and the setup steps vary from tool to tool. Some call it custom instructions, some organize it into projects or spaces, some let you save named skills. The idea underneath is the same everywhere: saved instructions the AI applies for you. Because the exact steps depend on which tool you use, we keep them in one place instead of scattering them here. See Module M1 for what this feature is called in the major tools and where to look, then follow your tool’s own help pages for the exact steps.

Think About It Think about one thing you explain to your AI tool over and over: some piece of background you retype most times you sit down. Who you are, what you’re working on, how you like the output. What is it? That answer is the seed of your first skill.

Build your first skill

Let’s build one together, then you’ll build your own.

Meet Marcus, a volunteer soccer coach. The thing he writes most is messages to his team’s families: game reminders, weather calls, snack schedules, the occasional “great effort today” note. Every one of them needs the same background. So Marcus writes a “team communications” skill. Here’s what he puts in it.

Who and what. “I coach the Northside U12 soccer team, twenty kids. My messages go to their parents. We practice Tuesdays and Thursdays at Riverside Park, field 3.”

Tone. “Warm and clear. I’m a volunteer, not a drill sergeant. Assume parents are busy and skimming on their phones.”

Format defaults. “Keep messages under 120 words. Lead with the single most important thing: a time, a date, a change. End with the field address when there’s a game or practice.”

Always and never. “Always give a specific makeup plan when I cancel, so parents aren’t left guessing. Never guilt anyone about attendance. Never use more than one exclamation point.”

That’s the whole skill. Four short chunks. Now when Marcus opens his tool and types “practice moved to Wednesday this week, field 2,” the reply comes back short, warm, parent-friendly, with the address and no guilt trip. He didn’t explain any of it. The skill did.

Notice the shape, because yours will use it too. A skill answers four questions: who you are and what you’re working on, what tone you want, what the output should look like, and the hard rules that never bend. Fill those four in for any task you repeat, and you have a skill.

Try It Now Pick the most recurring task in your own life: the thing from the “Think About It” a moment ago, or whatever you write most. Draft a skill for it using Marcus’s four parts: who and what, tone, format defaults, always and never. Keep each part to a sentence or two. Then open Module M1, find the setup steps for your tool, and save it as a real skill. Test it: give the AI a bare-bones request for that task and see how much it already knows. What did it get right without being told?

Prompt the prompt: meta-prompting

Here’s a move that feels like cheating. Instead of asking the AI to answer your question, first ask it to improve your question.

This is called meta-prompting, using the AI to make your prompt better before it answers. The instruction is short. Paste your prompt and add: “Improve this prompt first, show me the improved version, then answer it.” The tool rewrites your request, shows you the upgraded version, and only then does the work.

The magic is in what it adds. Watch the improved version closely. Nearly every time, it fills in the same things the cheat sheet told you a strong prompt needs: a sharper task, the context you left out, a clear format, a tone. You’ll see your one-line request grow the missing pieces in front of you. That’s not a coincidence. The AI is applying the same fundamentals you’re learning, which makes the improved prompt a free lesson every time you use it.

Use this when a request matters and you’re not sure how to frame it. Over a few weeks, you’ll notice the gap between your prompt and the improved one shrinking, because you start writing in the details it used to add for you.

Think About It Run meta-prompting on a real request this week. Set the improved version next to your original. What did it add: context, format, tone, or a constraint? Now check it against the cheat sheet’s “must include” list. How many of the things it added were things the cheat sheet already predicted you’d need?

The self-check habit

One more habit, and it’s the one that keeps you safe.

Before you accept any answer that involves facts, dates, numbers, or anything with real stakes, add this line: “Before finalizing, review your answer for errors or anything I should verify.” The AI reads back over its own work and flags the shaky parts: a statistic it’s unsure about, a date worth confirming, a claim you should check.

This connects straight back to what you learned about verifying AI output in Chapter 3. The self-check doesn’t replace your judgment; it points you at where to aim it. When the AI flags something, verify it. And stay alert to what it doesn’t flag, because it can be confidently wrong. The self-check is a smoke detector, not a fireproof house.

Keep the frame from the whole course in mind. The AI drafts. You judge, verify, and decide. Skills and libraries make the drafting faster and more consistent, but they don’t move the final call. You remain the editor of record. That never changes, no matter how good your setup gets.

What people get wrong here

The mistake isn’t using skills. It’s forgetting them.

A skill runs quietly in the background, which is exactly what makes it useful and exactly what makes it risky. Once it’s saved, you stop seeing it. So when the facts change and the skill doesn’t, it keeps applying old information to every answer, and you may not notice for weeks. Marcus’s skill still says “Riverside Park, field 3.” The team moved to a new field in the spring. Now every message he sends confidently gives parents the wrong address, and because the reply looks polished, nobody questions it, least of all Marcus.

A stale skill doesn’t fail loudly. It degrades quietly. Every answer gets a little bit wrong in the same invisible way, and the polish hides the rot.

The fix is a habit, not a tool. Every so often (a season change, the start of a school term, the first of the quarter), open your skills and read them like a stranger would. Is any of this out of date? A name, a date, a rule, a preference that shifted? Update it and move on. It takes two minutes.

You’ve actually seen this exact idea already. This course marks some of its own content as “perishable” and refreshes it on a schedule, precisely because facts go stale while the words on the page look just as confident as the day they were written. Your skills are the same. They’re only as current as the last time you looked.

Your move

Two things to build before the next chapter, and you already started one of them.

First, your prompt library. You created it in the first lab with a single entry. Add two more this week from tasks you actually do. Three good prompts is a real library.

Second, your first skill. You drafted one and saved it in Module M1. Use it at least three times this week on real work, and after the third time, go back and fix whatever it got wrong. That editing pass is the whole point.

Bring both to the capstone, the final chapter. You’ll put them to work on a full project from your own life, start to finish. Show up with a library you trust and a skill that sounds like you, and the capstone stops being an assignment and starts being the way you actually work.

This chapter was developed with AI assistance and reviewed by a human editor. It’s educational, fact-checked where applicable, and may contain minor errors. It’s not a substitute for professional advice.