Guide · for Codex or Claude Code
One folder for the whole course.
Keep every handout, lab, homework and note for CPSC 1710 in one folder, and open that folder with your AI coding agent. Then the agent can see the whole course, and so can you.
Why do this?
Several of you said the course material is spread over too many places.
Canvas, the labs site, Google Docs, Colab and GitHub each hold a piece. An AI coding agent reads the files in the folder you open it in. If everything is in that folder, you can ask it where a topic was covered, have it explain a file at your level, or have it quiz you before the midterm.
The setup takes about five minutes. You paste one prompt and your agent builds the folder. It also installs grill-me, a practice partner for the midterm that keeps notes on what you have covered.
Either agent works, and you can switch between them. Both read the same files.
Make an empty folder
Create a new folder called cpsc1710, for example inside Documents. Start with it empty. You will add your own files in step 4.
Open the folder in your agent
- Desktop app (Codex or Claude Code): start a new session and choose the
cpsc1710folder as the place to work. - VS Code: open the folder with File → Open Folder, then open the Codex or Claude Code panel.
- Terminal: go into the folder with
cd, then runcodexorclaude.
Check that the agent is working in cpsc1710 and not in your home folder. You can ask it: "Which folder are you working in?"
Paste this prompt
Your agent will ask for permission to download from github.com. Say yes. It then makes the folders, saves a copy of the labs site, installs the practice skill, and adds a short instruction file for itself. When it finishes, your folder looks like this:
cpsc1710/ ├── AGENTS.md how the agent should behave here, including the course's AI rules ├── README.md a map of the folder, for you ├── course/ │ ├── labs/ a copy of the labs site: lab pages, notebooks, practice questions │ ├── handouts/ empty: for files you download from Canvas │ └── slides/ empty: for lecture slides ├── work/ your own homework and lab work ├── notes/ your notes, and study notes the agent writes for you └── projects/ final project and experiments
If you use both agents, you only need to do this once. The folder works for both.
Add your own files, and keep everything
| What | Where it goes |
|---|---|
| Homework handouts from Canvas | course/handouts/ Download each one as a PDF. |
| Lecture slides | course/slides/ |
| Your homework and code | work/ One folder for each assignment, such as work/hw4/. |
| Your notes | notes/ Any format. |
Do not tidy by deleting. Old files cost nothing, and they are what lets the agent answer "where did we see this before?"
When new lab material is posted, say this to your agent:
Ask across the whole course
Try one of these. Change the topic to something you are unsure about.
Practise for the midterm with grill-me
The midterm is on paper. You may bring notes on paper, but no devices and no AI. The best preparation is answering questions, and grill-me asks them. Start it by saying what you want:
You can also type $grill-me in Codex or /grill-me in Claude Code. If the skill does not appear, close the app and open it again.
- One question at a time. It waits for your answer before it says anything else.
- It adjusts to you. Get two right and the questions get harder. Miss one and it stays on that idea, then brings it back later.
- It uses the exam's formats: multiple choice, ordering, a small calculation by hand, pseudocode, short explanations.
- It keeps notes in
notes/grill-me/: your progress by topic, a list of what you missed with the right idea next to each, and a log of every session. The next session starts from those notes, in either agent.
Other things to ask for:
Three habits that make it work. Write by-hand and pseudocode answers on paper before you type them. Do the course's own practice sets yourself, on paper, before asking the agent about them. Say "I don't know" when you don't: it will give a hint, not the answer.
For the curious
What was installed
grill-me on GitHub
A skill is a short instruction file that an agent loads when a task calls for it. This one is a page of plain English. Read it, and change your copy if you want it to behave differently.
Read the skill → The setup instructionsWhat the prompt tells your agent to do
The setup prompt points your agent at this file. It is written for the agent, and it is short enough to read before you run it.
Read the instructions →