Week 3

Idea to production

Prompting as a skill

Debugging with the agent

Security essentials

Practice

Assignment

Back to Track

Watch: The official short version of this week's flow. Ours adds polish, tests and review on top.

https://www.youtube.com/watch?v=xJQuF02NAK8&list=PLmWCw1CzcFilebjK89WLb5cAvM8K0cLB3&index=6

Every feature travels the same path. Naming the stages matters because what you delegate to the agent and what you verify by hand is different at each one.

flow-diagram.png

Before anything: define "done"

Write one or two sentences before the first prompt: what exactly will exist when this is finished, and how will I check it? "A 'Member since' date on the profile page, formatted like the rest of the card, visible for logged-in users." Without this, beginners fall into one of two traps: polishing endlessly, or merging something raw. "Done" defined up front is the exit condition for both.

The stages

1. Skeleton. Structure and stubs, no real logic. Delegate freely — the agent is excellent at scaffolding. Verify: does the structure match how the project already organizes things (Week 2)?

2. Fill-in. The real logic. Delegate in small units, one behavior at a time. Verify: read every diff; run the thing after each step. This is where most debugging happens — see the debugging page.

3. Polish. Edge cases, formatting, naming. Delegate the mechanical parts ("handle the empty state everywhere this pattern appears"). Verify: against your "done" definition — not against perfection.

4. Testing. Ask the agent to propose test cases first, then write tests — reviewing a list of cases is easier than reviewing test code. Verify: do the tests test real behavior, or just mirror the implementation? Run them; break something on purpose and check a test catches it.

5. PR. One branch, a reviewable diff, an honest description (what the agent did, what you changed, what you rejected — the Week 2 format). Before pushing: the security checklist from this week's security page.

6. Review. Fresh eyes on the result — a separate review chat (Week 1 habit), a classmate, or a mentor. Respond to feedback in small commits.

When to stop

Stop when "done" is met and tests pass — not when you run out of improvement ideas. The agent will always have another suggestion; that's not a reason to take it. Park extras in a follow-up note.


The HackYourFuture curriculum is licensed under CC BY-NC-SA 4.0 *https://hackyourfuture.net/*

CC BY-NC-SA 4.0 Icons

Built with ❤️ by the HackYourFuture community · Thank you, contributors

Found a mistake or have a suggestion? Let us know in the feedback form.