Free downloads

Five prompts from class, with the explanation we give in the room.

Each one is a plain text file. Download it, paste the prompt into ChatGPT, Claude or Gemini, fill in the brackets, and read the "watch out for" part before you trust the output. No account, no email, no catch.

Match your resume to a job post

Like you are ten: you found a job on Indeed. This prompt has AI read the job post and your resume side by side, then reword your bullets so the hiring manager can see the match. It also lists the keywords the company's resume-reading computer looks for. Your experience stays the same. The wording does the traveling.

How: copy the job post, copy your resume, paste both into the prompt, read every line it gives back. If it wrote something you have not done, delete it. The prompt itself orders it not to invent anything, but you are the fact-checker.

Watch out for: AI smooths things over to be helpful. A resume that exaggerates gets found out at the interview. And leave your phone number, address and anything like a Social Security number out of what you paste.

Download Text file · works in ChatGPT, Claude, Gemini

Make a piece of content, start to finish

Like you are ten: you have an idea for a post or a video. This prompt turns AI into a writing partner that gives you three different openings, a full draft built from YOUR notes, and three short versions for social media. It follows your voice rules, so the draft sounds like you instead of like the internet.

How: one sentence for the idea, three to five points you want to make, two or three rules about how you talk. Paste it all in. Then cut at least a tenth of what comes back, because first drafts always carry filler.

Watch out for: AI drafts sound generic until you swap its examples for your real stories. Your story from last Tuesday beats anything it invents.

Download Text file · works in ChatGPT, Claude, Gemini

Set up an "AI employee" for one repeat task

Like you are ten: an "AI employee" is a saved job description for one task you do over and over, like replying to routine emails or summarizing long documents. You write the instructions once, save them, and paste them at the start of each new chat instead of re-explaining the job every time.

How: pick ONE repeat task, fill in the template like you are training a new hire on their first day, save it in your notes app, and check its work every time at first. Honest version: it is not an employee. It is a checklist that talks. It does not remember you between chats and it does not know when it is wrong.

Watch out for: every mistake it makes twice becomes a new line in the rules. That is the training. And read anything with facts, money or names before it leaves your hands.

Want the developer version? The same idea exists as installable files for coding tools: agency-agents on GitHub (a big shelf of ready-made roles, free) and Anthropic's official skills repo. Those need a coding agent like Claude Code or Cursor. The download here needs nothing but a browser.

Download Text file · works in ChatGPT, Claude, Gemini

Find your direction (the four-questions exercise)

Like you are ten: this one is for the big question, "what should I be doing?" It comes from the Japanese idea of ikigai: the sweet spot where four things overlap. What you love, what you are good at, what people need, and what people pay for. AI plays the patient interviewer, asks you twelve plain questions one at a time, then compresses your answers into one sentence you can steer by.

How: set aside twenty minutes, answer in your own words, and if the sentence it gives you could describe a hundred other people, say so and run it again with narrower answers. Works for picking a side business, a channel, or thinking about a career move.

Watch out for: it is a mirror, not a fortune teller, and not career advice. It organizes what you already know about yourself. And the exercise ends with a small two-week test for a reason: test small before you leap.

Download Text file · works in ChatGPT, Claude, Gemini

Start a second brain for your notes and ideas

Like you are ten: a second brain is one place where your ideas and things-to-remember live, organized so future-you can find them. This prompt has AI interview you about how you work, then design the smallest folder setup that fits, in whatever notes app you already have.

How: answer its five questions straight, especially the one about systems you tried and dropped. Then set up exactly what it suggests, nothing more. The trap with second brains is building a beautiful system you abandon in two weeks, and this prompt is built to keep yours small enough to survive.

Watch out for: the weekly ten-minute review IS the system. Skip it for a month and you are back to scattered notes. And do not buy a new app for this. The habit matters, the app does not.

Download Text file · works in ChatGPT, Claude, Gemini
For the more advanced crowd

The same idea, installed: AI employees from GitHub.

Everything above works by pasting into a chat window. There is a bigger version of the "AI employee" idea for people comfortable with developer tools, and it is free and open source.

What it is

A GitHub project called agency-agents ships 144 ready-made "employees": a content creator, a growth hacker, a project manager, a security engineer and more. Each one is a Markdown file, a longer, more disciplined version of the job-description prompt in our download above: the role, the standards, the output format, what it must refuse to do.

How it differs from pasting

Instead of pasting instructions into a chat each time, you install the files into a coding agent such as Claude Code, Cursor or Codex. The agent loads the right employee when the task matches, remembers your project between sessions, and can read your files and run commands. Pasting is a conversation. Installing is hiring.

How to try it

Clone the repo, open the folder for one role that matches work you do, and read it before installing anything. One command to see everything: git clone https://github.com/msitarzewski/agency-agents.git. It is MIT licensed and actively maintained. Start with one employee doing one real task, and review its work like a new hire's first week.

Honest limits

An installed employee is still not staff. It is fast, tireless and confident, and confidence is exactly why you check its work. Give it one job, watch the first outputs closely, and keep anything with money, credentials or other people's data in human hands.

Want to learn this in a room instead of from a file?

Every one of these started as a class exercise. We teach them live, free, at libraries across Westchester, Putnam, Dutchess and the Harlem Valley. Bring your own problem and we will work on it together.