Working with medical imaging data is hard. DICOM is a complex standard, DICOMweb has its quirks, and building reliable imaging pipelines requires deep domain knowledge that most developers don't carry around in their heads.
Today we're releasing Aurabox Skills, a set of open source skills for AI coding agents that bring medical imaging expertise directly into your development environment.
What are skills?
Skills are structured knowledge files that AI coding agents (like Claude Code) load automatically when they're relevant to what you're working on. They give your agent deep, domain-specific context without you having to explain the intricacies of DICOM tags or multipart MIME encoding every time you start a new task.
When a skill isn't needed, it has zero cost. Only the description is loaded at startup. When your prompt matches a skill's domain, the full reference material is pulled into context automatically.
What's included
We're releasing four skills covering the most common areas where developers need help with medical imaging.
Aurabox REST API
A complete API reference for developers integrating with the Aurabox platform. Includes working client code in Python, TypeScript, and curl for managing patients, cases, and studies, along with authentication, pagination, and error handling patterns.
DICOM Processing
Everything you need to read, write, and manipulate DICOM files. Covers pydicom fundamentals, a tag reference table with common tags and their value representations, pixel data access and windowing, transfer syntax handling, and bulk operations. Also includes DCMTK command-line tool references for when you need to work outside of Python.
DICOMweb Protocol
A practical guide to building HTTP-based imaging integrations. Covers QIDO-RS for searching, WADO-RS for retrieval, and STOW-RS for uploads (including the multipart MIME encoding that trips up most developers). Includes a complete DICOMwebClient class and server-specific notes for Aurabox, Orthanc, dcm4chee, and Google Cloud Healthcare.
Medical Imaging Pipelines
For researchers and ML engineers working with imaging datasets. Covers format conversion (DICOM to NIfTI, PNG, JPEG, HDF5), CT and MRI intensity normalisation, resampling and preprocessing, metadata manifest generation, and a complete CT research pipeline example with PyTorch Dataset integration.
Getting started
The quickest way to install skills is with the skills CLI:
npx skills add https://github.com/aurabx/skills
This will let you select which skills to install and where to put them.
You can also install skills manually by copying them into your project's .claude/skills/ directory or your personal ~/.claude/skills/ directory:
# Clone the repository
git clone https://github.com/aurabx/skills.git
# Install a skill to your project
cp -r skills/skills/dicom-processing /path/to/your/project/.claude/skills/
# Or install globally for all projects
cp -r skills/skills/dicom-processing ~/.claude/skills/
Restart your agent after installing. Skills trigger automatically based on your prompts -- no manual invocation needed.
Why we built this
Medical imaging software is critical infrastructure. The people building it should have access to the best tools and references available, regardless of whether they use Aurabox.
These skills encode the kind of knowledge that usually lives in a senior engineer's head or scattered across specification documents. By packaging it as agent skills, we can put that expertise within reach of every developer working in this space.
Contributing
The skills are open source and available at github.com/aurabx/skills. If you find an error, want to expand a reference, or have ideas for new skills, contributions are welcome.
We hope these skills make medical imaging development a little less painful. If you have questions or feedback, reach out to us on GitHub or contact our team directly.