Skillnaija Start free

An agent pack for one real chore

The capstone project for Agentic AI: Build and Supervise AI Agents That Take Action, Safely — a free AI & Emerging Tech course. Pass it and you earn a certificate anyone can verify.

Start the course free Download the project workbook

The project unlocks once you complete the lessons. The workbook is free to download now so you can see exactly what is expected.

What you will submit

One shared Google Doc link with the chore brief and decision, the loop trace and tool list, evidence of a working build with its knowledge pack and approval sheet, the permission ladder and stop plan, the risk check, the test results with log rows and a cost estimate, and the rules note.

What to do, step by step

  1. Choose one real business (your own, a family business, an employer or a client; if you have none, a local business you know well, and say so) and one repetitive chore in it. Everything you build, test and submit uses made-up data: invented customers, invoice numbers and amounts. Never include anyone's real personal data; never give an agent a real banking login, payment key, password, BVN, NIN or OTP; and never let any key or password, including your model API key, appear in your document or in a screenshot inside it (the link is open to anyone who has it).
  2. Part 1. Chore brief and decision (one page): the chore, who does it now and how many hours a week it takes, what 'done' looks like, and your decision for each piece of the work (plain workflow, workflow with one AI step, or agent) with one sentence of reasoning for each piece.
  3. Part 2. The loop and the tools: the agent's goal (the job, what done means, what it must never do, when it hands over), one made-up case traced turn by turn (plan, action, what came back) with a turn limit, and the tool list. For each tool say whether it is read or write and finish the sentence 'This tool lets the agent ___, and it must NOT be able to ___'. List the tools you deliberately did not give it.
  4. Part 3. The build: a working agent on a free tier, either the no-code track (Lesson 5) or the laptop track (Lesson 6), run on at least ten made-up cases. Write down the result you expect for each case before you run it: these ten cases are also your Part 6 test set. Show the instructions you gave it, screenshots or a log of two runs, the knowledge pack it uses (each item, pasted or looked up, its owner and a last-reviewed date), and the approval sheet with at least five drafts and your decision on each (approved, edited or rejected, with a reason). The agent itself sends nothing.
  5. Part 4. Permissions and stop plan: a permission ladder that places every action on a rung (read, reversible write, never alone) and lists what is not connected at all, who approves what and when, the five things your agent never does alone (send, pay, delete, accept terms, submit), and how a non-technical owner pauses it in under a minute.
  6. Part 5. Risk check: a table marking, for each piece, whether it reads untrusted content, holds private data and can change records or act outside the business; one harmless injection test on your own build (the made-up line you used, what the draft said, what the agent could and could not have reached); and the design change you made or would make.
  7. Part 6. Tests, log and cost: your test set of at least ten made-up cases (the ten from Part 3 will do) with a pass rule for each, every case run three times with pass or fail recorded, what you changed after the first round, and the result of running the set once more; ten log rows with no personal data (on the no-code track, copied by hand from the builder's record of each run); and one month's cost estimate with the working shown and the date you read the live page, in the provider's own currency: on the laptop track's free plan, the usage reading before and after your runs as the tool's usage command shows it, with a note that the free plan shows no money figure (multiply tokens by the model's price only if you paid for API credit); or, on the no-code track, runs multiplied by the builder's own unit (activities, credits, executions or tasks) set against the free monthly allowance and the price of the next plan up.
  8. Part 7. Rules note (half a page): what personal data the chore involves and how you kept it out or minimised it, where data would be sent, which decisions stay with a named human, and three questions you would put to the owner, a lawyer or a data protection officer before real data is used. Say which of the rules you mention are law, which are a provider's terms and which are drafts or voluntary guidance.
  9. Put everything in one Google Doc, set sharing to 'anyone with the link can view', and submit the link.

Files to work with

Project workbook (fill in, then submit)The whole brief, an evidence checklist, the grading rubric as a self-check and the link-sharing steps in one file. Opens in Word, Google Docs or WPS.Word · 8 KB

How it is graded

CriterionWeight
The chore is real and specific, each piece has a justified workflow-or-agent decision, and the goal and traced loop show a task an agent can actually finish and stop20%
The tools are scoped narrowly, the build works on made-up data, and every outbound, money or irreversible action sits behind a human approval that the evidence shows was really used25%
The permission ladder, the five never-alone lines, the risk table and the harmless injection test show that limits were designed in, not assumed20%
The test set has clear pass rules and repeat runs, the log is readable and free of personal data, and the cost estimate shows its working (in tokens or in the builder's own unit) in the provider's currency with a date20%
The rules note keeps personal data out or minimised, names the human who decides, separates law from provider terms and drafts, and the submission contains no real personal data, logins or keys15%

You need 65% overall to pass. A failed submission comes back with feedback and can be revised and resubmitted once free; further attempts come with the All-Access Pass or as a paid retake.

Before you submit: make your link public

If your work lives in Google Drive or Google Docs, open Share → General access and change “Restricted” to “Anyone with the link” (Viewer). Then paste the link into a private browser window: if it opens without sign-in, you are ready. A private link cannot be graded — it is the single most common reason a good project fails.

Frequently asked questions

Do I have to finish Agentic AI: Build and Supervise AI Agents That Take Action, Safely before submitting the project?

Yes. The submission screen opens once every lesson in Agentic AI: Build and Supervise AI Agents That Take Action, Safely is marked complete. The lessons are where the methods, the local context and the worked examples the project depends on are taught.

How is the project graded?

An examiner scores each rubric criterion from 0 to 100 and weights them as shown on this page; you need 65% overall to pass. Most submissions are graded within minutes and you get written feedback on what was strong and what to improve.

Can I resubmit if I fail?

Yes. A failed project comes back with feedback; revise the weak parts and resubmit. A passed project is final — the certificate is issued and cannot be re-rolled.

What do I actually submit?

A write-up of what you did (under 5,000 characters) plus a public link to your work — a Google Drive folder, Google Doc, spreadsheet, GitHub repository or video. The link must open without sign-in; a private link cannot be graded. The free project workbook on this page walks you through all 9 steps.

Is the certificate real?

Yes. Passing this project issues a certificate with a unique verification code. Anyone — an employer, a client, a school — can open the verification page and see that the certificate is genuine and which project earned it.

More AI & Emerging Tech projects

Ready to earn this certificate?

Agentic AI: Build and Supervise AI Agents That Take Action, Safely is self-paced. Finish the lessons, complete this project, and the certificate is yours to share.

Start learning free