dataset-curator

VEuPathDB Dataset Curator

Claude Skills for curating new datasets for VEuPathDB resources with AI assistance by Claude Code.

For Curators

Prerequisites

Install Volta

Volta manages Node.js versions:

curl https://get.volta.sh | bash

Install Node.js

volta install node

Install Claude Code

volta install @anthropic-ai/claude-code

Note: The first time you run claude, you'll need to complete a one-time account linking and login process.

GitHub Desktop (Recommended)

If you don't already have it, download from desktop.github.com

Setup

Install the Curation Skills Plugin

Start Claude Code with claude from any directory. However, it's a good idea to make a dedicated directory to do Claude-based curation work and run claude from inside that directory. (See "Starting a Curation Session" below.)

1. Add the VEuPathDB marketplace to Claude Code:

   /plugin marketplace add VEuPathDB/dataset-curator
   

   It will do some installation work and should hopefully report back "Successfully added marketplace: dataset-curator".

2. Install the curation-skills plugin:

   Re-enter the plugin management interface (still in Claude Code)

   /plugin
   

   • Choose Option 1
   • Use Space to select the curation-skills plugin
   • Press i to install
   • Restart Claude Code (/exit then run claude again) to load the skills

That's it for setup! You're ready to start curating.

Note on VEuPathDB Repositories: Skills need access to one or more VEuPathDB configuration repositories (e.g. ApiCommonDatasets, ApiCommonPresenters, EbrcModelCommon). If you already have these cloned via GitHub Desktop, you'll create a symlink to them from your curation workspace directory. If not, the skill will guide you through setting them up when you first run it.

Usage

Starting a Curation Session

1. Create a curation workspace directory for your curation sessions:

   mkdir ~/my-curation-workspace
   cd ~/my-curation-workspace
   

   You can create different workspace directories for different datasets, or reuse the same one.

2. Set up a link to your GitHub repositories in your curation workspace directory:

   ln -s ~/Documents/GitHub veupathdb-repos
   

   This links to your GitHub Desktop repositories so changes appear in your actual clones. Claude Code will edit files in these repositories, but you will manage branching, committing, pulling and pushing using the GitHub Desktop GUI.

3. Start Claude Code:

   claude
   

4. Tell Claude what you want to do:

   I want to curate a new genome assembly
   

Claude will activate the appropriate skill and guide you through the workflow.

Important: Follow the git branching guidelines before starting. Create dataset-specific branches in your repositories using GitHub Desktop.

What Happens During Curation

• Claude Code handles: Fetching NCBI data, processing metadata, generating XML configurations, updating files
• You handle: Git operations (branches, commits, pull requests) via GitHub Desktop or command line

This separation ensures you maintain full control of your git history and can easily review or rollback changes.

Updating Claude Code

Claude Code is under active development with frequent updates and improvements. Although Claude Code displays "Auto-updating..." on startup, this mechanism doesn't work reliably when installed via Volta.

Recommended: Manually update Claude Code weekly by rerunning the installation command:

volta install @anthropic-ai/claude-code

This ensures you have the latest features, bug fixes, and improvements.

Updating Skills

As we improve and fix bugs in the curation skills, you'll want to update to the latest version:

1. Check for updates using the plugin management UI:

   /plugin
   

   • Select 2. Manage and uninstall plugins
   • Select dataset-curator marketplace
   • Select curation-skills plugin (<Enter> for details)
   • Choose Update now if an update is available

2. Restart Claude Code sessions: Close any active claude sessions and start fresh. Skills are loaded when Claude Code starts.

Important: Don't update skills in the middle of a curation session. Complete your current workflow, commit your changes, then update skills before starting a new dataset.

Available Skills

• curate-genome-assembly: Process genome assembly datasets - fetch NCBI metadata, generate organism XML, update ApiCommonDatasets configurations
• curate-bulk-rnaseq: Process bulk RNA-seq datasets - fetch SRA/GEO metadata, analyze sample factors, generate presenter XML and pipeline configurations

---

For Developers