Day 1 Challenge

This challenge ties the whole day together into a single piece of proof. You’ll capture a fingerprint of your laptop, carry it across to the Yens with scp, and then hand the analysis and the git work to Claude Code — the tool you just set up. The result is one file, day1_challenge.json, that travels from your laptop to the cluster and lands in your fork as a pull request.

Everything you’ll use today shows up here: SSH, scp, Claude Code, and the fork → branch → commit → push → PR workflow from the github-for-research skill.

The Challenge

Task: Capture your laptop’s public IP, scp it to the Yens, then have Claude Code compute a letter distribution over the grimoire spell names, write both into day1_challenge.json, and open a pull request on your fork.

The finished file has this structure (the numbers below are placeholders to show the shape — Claude computes the real counts):

{
  "public_ip": "203.0.113.42",
  "letter_distribution": { "a": 0, "b": 0, "c": 0 }
}

public_ip — the public IP address of your laptop (proof the file started on your machine and crossed to the cluster).
letter_distribution — how often each letter a–z appears across the spell name field of the grimoire (Claude computes this for you).

Step 1 — Grab your laptop’s public IP

Do this on your laptop (a local Terminal / Git Bash tab), not on the Yens — run it on the Yens and you’d capture the cluster’s address instead of yours.

curl -s https://api.ipify.org > ~/Desktop/provenance.txt   # the public IP of your laptop
cat ~/Desktop/provenance.txt                               # confirm you got an address

You should see something like 203.0.113.42.

Everyone in the room shares the same network, so this address points at the classroom, not at you personally.

Step 2 — Send it to the Yens with `scp`

Still on your laptop, copy the file straight into your course repo on the Yens (scp runs over the SSH connection you set up earlier):

scp ~/Desktop/provenance.txt SUNetID@yen.stanford.edu:~/gsb-research-computing-ai-skills/provenance.txt

This is the same upload direction you practiced in Transferring Files (scp) — source is local, destination is remote.

Step 3 — Let Claude Code do the analysis and the git work

Now hop onto the Yens, move into your repo, and launch Claude Code:

ssh SUNetID@yen.stanford.edu     # if not already connected
cd ~/gsb-research-computing-ai-skills
ml claude-code
claude

Press Shift+Tab until you reach plan mode — so Claude investigates and proposes an approach without changing anything until you approve. This is the look → plan → approve → act loop from Working with Claude Code.

Now describe the task in your own words — don’t copy a script. Tell Claude what you want to end up with, and let it work out the how. Your instruction should cover all of these objectives:

Analyze the grimoire. The dataset is already in the repo you cloned, at docs/assets/data/grimoire.zip — Claude just needs to unzip it (no download). Count how often each letter a–z appears across the spell name field only — the first underscore-separated field of each filename (e.g. ashstrike in ashstrike_fire_3_defensive_meteor.spell), never the element, tier, type, or mastery. Be explicit about how you count: use every one of the files’ names (names repeat across the dataset, and each file counts), lowercase a–z only, as integer counts.
Read your provenance. Pull your laptop’s public IP from provenance.txt, the file you just scp‘d into this repo.
Write the result. Produce a file named day1_challenge.json in the repo, containing exactly two top-level keys and nothing else:
- public_ip — the IP string from provenance.txt.
- letter_distribution — an object mapping each lowercase letter to its integer count.
The file’s structure must match this (the numbers are placeholders — Claude fills in the real counts):
```
{
  "public_ip": "203.0.113.42",
  "letter_distribution": { "a": 0, "b": 0, "c": 0 }
}
```
Do the git work the research way. Log an issue for the task, work on a new branch, commit crediting Claude, and open a pull request. (Logging the issue needs the Issues tab switched on for your fork — Step 1 of Version Control with Git. If Claude reports that it can’t create an issue, that’s why.)

Then read the plan Claude gives you. Before you approve, check it yourself:

Does it hit every objective above — the analysis, the IP, the JSON file, and the full issue → branch → commit → PR cycle?
Is it following the github-for-research skill you installed in Working with Claude Code? If the plan doesn’t mention the skill or the branch/issue/PR habits, ask Claude: > are you using the github-for-research skill for this? — and have it revise the plan if not.

Refine the plan until it clearly achieves the objectives, then approve it and let Claude carry it out. Notice what you didn’t do: no curl for the dataset, no unzip, no counting by hand, no git commands typed out — you described the outcome and reviewed the approach, and Claude did the rest.

Step 4 — Verify your submission

Open your fork on GitHub in a browser. You should see a new pull request with day1_challenge.json in its Files changed tab. Open the file and check it has both keys — a valid IP in public_ip and a set of letter counts in letter_distribution. That pull request is your submission.

Merging it is optional — the open PR is enough to count as done.

If Claude reports that git push failed for authentication, you haven’t signed in to GitHub on the Yens yet. If you still haven’t authenticated, follow Step 3 of Version Control with Git to set up your token — or ask an instructor for help. Then ask Claude to push again.

🟢 Green sticky = I’m done and ready 🔴 Red sticky = I need help

Put a sticky note on your laptop lid so instructors can see where you are.

Day 1 Challenge complete

What You Learned

Skill	Where you learned it
`curl` to fetch data over the network	Bulk File Operations
`ssh` to a remote server	Connecting to a Cluster
`scp` to move a file laptop → cluster	Transferring Files (scp)
The spell filename fields (`name_element_tier_type_mastery`)	Bulk File Operations
Driving a task in plain English with Claude Code	Working with Claude Code
Branch → commit → PR, done for you by the skill	Version Control with Git + Working with Claude Code

You just ran a full research loop end to end: fingerprint your machine, move data across the network, analyze it with an AI assistant, and version-control the result — the same shape of workflow you’ll use for real projects on the Yens.