
scripps-cbb/scripps-hackathon-starter
Developer toolsA ready-to-fork template for participants of the (May 29 – June 1, La Jolla).
Overview
A ready-to-fork template for participants of the (May 29 – June 1, La Jolla). - A workspace pre-configured with two specialized skills: - /hackathon-aws — full context on the hackathon's AWS account (VPC, EC2, S3, EICE SSH, Bedrock) - scripps-garibaldi-hpc — Garibaldi cluster conventions, Slurm templates, GPU job patterns - for both platforms (AWS_SETUP.md and HPC_SETUP.md) - (hello-world/) that get you fluent with Claude Code (slash commands + auto mode) and then validate your HPC and AWS accounts Welcome! This page is the homepage of a — a folder of files tracked by Git. You don't need to learn Git to use this template. The fastest path: 1. Pick whichever path fits you: - Sign in to GitHub, click at the top right of the repo page, then clone your fork — e.g. git clone https://github.com/ /scripps-hackathon-starter. This gives you a personal copy you can commit to all weekend, share with teammates, and revisit after the hackathon.
README
Scripps Hackathon Starter
A ready-to-fork template for participants of the Scripps Research 2026 Computational Biology & Bioinformatics Hackathon (May 29 – June 1, La Jolla).
It gives you, out of the box:
- A Claude Code workspace pre-configured with two specialized skills:
/hackathon-aws— full context on the hackathon’s AWS account (VPC, EC2, S3, EICE SSH, Bedrock)scripps-garibaldi-hpc— Garibaldi cluster conventions, Slurm templates, GPU job patterns
- Setup guides for both platforms (AWS_SETUP.md and HPC_SETUP.md)
- Three “Hello World” exercises (hello-world/) that get you fluent with Claude Code (slash commands + auto mode) and then validate your HPC and AWS accounts
Never used GitHub before? Start here.
Welcome! This page is the homepage of a repository — a folder of files tracked by Git. You don’t need to learn Git to use this template. The fastest path:
-
Get the code on your laptop. Pick whichever path fits you:
- Fork (recommended if you want to keep your work). Sign in to
GitHub, click Fork at the top right of the repo page, then
clone your fork — e.g.
git clone https://github.com//scripps-hackathon-starter. This gives you a personal copy you can commit to all weekend, share with teammates, and revisit after the hackathon. Drop your project code in a new folder of your fork (e.g.my-project/) so it lives alongside the starter setup. - Clone read-only.
git clone https://github.com/scripps-cbb/scripps-hackathon-starter— quickest if you already have a git account but don’t plan to push code back. - Download ZIP (no GitHub account needed). Click the green
<> Codebutton → Download ZIP → unzip it somewhere you can find again. Lowest friction; you won’t be able to push your code back to GitHub from this copy.
- Fork (recommended if you want to keep your work). Sign in to
GitHub, click Fork at the top right of the repo page, then
clone your fork — e.g.
-
Install Claude Code if you haven’t already:
-
Open this folder in Claude Code. On Mac/Linux:
cd scripps-hackathon-starter && claude. On Windows: open the folder in VS Code with the Claude Code extension installed. -
Run the Hello World exercises below in order. They take ~15 minutes total and tell you exactly which piece (if any) of your setup needs attention.
You’ll spend almost all of your hackathon time talking to Claude in plain English — telling it what you want, and letting it run the commands. The skills in this repo are what make that work for AWS and HPC specifically.
Stuck? Open Claude Code in this folder and just ask. The skills teach Claude what’s in this repo, so questions like “what does this file do?” or “I got this error — what should I try?” work well.
What’s in this repo
.
├── README.md <- you are here
├── AWS_SETUP.md <- set up AWS CLI + SSO + the skill
├── HPC_SETUP.md <- set up Garibaldi HPC access
├── test_skill.sh <- end-to-end AWS smoke test (bash)
├── hello-world/
│ ├── README.md <- three-exercise walkthrough
│ ├── hello_claude.md <- "drive Claude Code" (slash cmds, modes)
│ ├── hello_hpc.slurm <- "does Slurm work?" (Garibaldi)
│ └── hello_aws.py <- bonus: "does AWS Bedrock work?" (Python)
└── .claude/
├── commands/
│ └── hackathon-aws.md <- the /hackathon-aws slash command
└── skills/
└── scripps-garibaldi-hpc/
└── SKILL.md <- auto-loaded HPC skill
Quickstart
TL;DR — one-prompt smoke test (for experienced users). Open Claude Code in any folder, paste the prompt below, and it’ll clone this repo and run the Hello World end-to-end. New to all this? Skip this box and follow the four numbered steps below instead — they explain what you’re doing as you go.
Clone https://github.com/scripps-cbb/scripps-hackathon-starter onto my Desktop, cd in, then run the Hello World as a smoke test: - Skip hello_claude.md (the Claude Code tour) — I already know the basics. - Lint hello-world/hello_hpc.slurm with `bash -n` and report any issues. Don't try to submit it. - Run ./test_skill.sh against my AWS SSO profile . If my token is expired, tell me to run `aws sso login --profile ` and wait for me to confirm before retrying. The script will create an S3 bucket, briefly launch a t3.micro, and terminate it — that's expected. - If you can pip install boto3, also run hello-world/hello_aws.py with my profile to confirm Bedrock works. - Stop and ask me before terminating or deleting anything I didn't create in this session. End with a punch list of any failures and what I should do next.Replace `` with your AWS SSO profile name (both places). No profile yet? Walk through AWS_SETUP.md first, then come back here.
1. Get fluent with Claude Code (5 minutes)
Open this repo in Claude Code and follow hello-world/hello_claude.md. You’ll try slash commands, switch into auto-accept-edits / plan mode, and run auto mode so you’ve seen all of it once before it matters.
2. Set up the HPC (5 minutes once your account is provisioned)
Follow HPC_SETUP.md. At the end you’ll sbatch hello_hpc.slurm and see your job run on a Garibaldi compute node.
3. Set up AWS (10 minutes, one-time)
Follow AWS_SETUP.md. At the end you’ll run:
./test_skill.sh
It creates an S3 bucket, reads a public dataset, launches and terminates a tiny EC2 instance, and tells you PASS/FAIL for each step.
4. Talk to Claude
Open this repo in Claude Code and try any of these:
/hackathon-aws launch a t3.medium and SSH into it
write me an sbatch script for Garibaldi that runs train.py on 1 GPU for 4 hours
create an S3 bucket called scrippsresearch--hackathon and upload my results folder
I got "Token has expired" — what now?
The first request hits the AWS slash command. The second auto-loads the HPC skill. The third works with either, depending on context. The fourth is just a normal conversation — Claude already knows what to do because the skills tell it.
Shared hackathon data
The organizers have pre-loaded project datasets into a shared S3
bucket: s3://scrippsresearch-hackathon/. Once your SSO is set up
(AWS_SETUP.md) and aws sso login has succeeded, you
can browse and download with your own profile — no extra credentials.
# See what's available
aws s3 ls s3://scrippsresearch-hackathon/ --profile
# Peek inside a specific project
aws s3 ls s3://scrippsresearch-hackathon/allen/ --profile
# Pull a project's data down to your laptop (or EC2 / HPC scratch)
mkdir -p data/allen
aws s3 sync s3://scrippsresearch-hackathon/allen/ data/allen/ \
--profile
Each prefix is one team’s project. Start by reading the project’s own docs for context, file conventions, and what the inputs actually mean:
| Prefix | Topic | Start here |
|---|---|---|
allen/ |
PQBP1 hexamer cryo-EM + docking | PQBP1_Hexamer_IM_README.md and _UPLOAD_NOTES.md inside the prefix |
heberling/ |
Siglec6 structural work | Hackathon Materials Notes JLH.pdf |
Quick way to read a project README without downloading the whole thing:
aws s3 cp s3://scrippsresearch-hackathon/allen/PQBP1_Hexamer_IM_README.md - \
--profile
(The trailing - streams to stdout.)
Big datasets? Copy only what you need.
aws s3 cp --recursiveandaws s3 syncboth accept--exclude '*'/--include 'pattern'so you can grab e.g. just the.pdbfiles. For 10+ GB downloads, consider running thesyncfrom an EC2 instance instead of your laptop — S3 → EC2 traffic inus-west-2is free via the VPC gateway endpoint.
How the Claude tooling works (in 60 seconds)
There are two kinds of context-providers in this repo:
- Slash commands live in
.claude/commands/. You invoke them explicitly:/hackathon-aws. Use this when you want to be sure Claude has the full AWS context loaded. - Skills live in
.claude/skills//SKILL.md. Claude loads them automatically when your question matches the skill’s description. You don’t type anything special — just ask about Garibaldi and the cluster skill kicks in.
Both files are plain Markdown — open them and read along if you’re curious. Editing them changes how Claude behaves in your session.
Personalize before you start
A few values in this template point at the maintainer’s defaults. Update them once and the rest of the toolchain follows:
| File | What to change |
|---|---|
| test_skill.sh | Set PROFILE to your SSO profile name, and BUCKET to a name that includes your username. |
| hello-world/hello_aws.py | Default --profile argument (or pass it on the CLI). |
| .claude/commands/hackathon-aws.md | The skill references inewman-wsl as the example profile — when you start a Claude session, just tell it “use profile `` for all AWS commands” and it will. No file edit needed. |
If you’d like Claude to remember your profile across sessions, ask it
to save a memory: “Remember that my AWS profile is bgood-scripps.”
Where to go from here
- Hackathon homepage and schedule:
- AWS account ID / SSO portal: see AWS_SETUP.md
- HPC contact:
[email protected](for Garibaldi account requests) - Claude Code docs:
Good luck and have fun. The whole point of the weekend is to try something you wouldn’t normally try — the tools in this repo are here so the infrastructure isn’t what slows you down.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add scripps-cbb/scripps-hackathon-starter