Aurabox Skills

Skills for working with the Aurabox medical imaging platform and medical imaging data.

---

🛠️ This project was built by Aurabox

Some of our other projects...

---

Skills

Aurabox API

Skill	Description
aurabox-rest-api	Client code for the Aurabox REST API (patients, cases, studies)

Medical Imaging

Skill	Description
dicom-processing	Read, write, and manipulate DICOM files with pydicom and DCMTK
dicomweb-protocol	Build DICOMweb clients (WADO-RS, STOW-RS, QIDO-RS)
medical-imaging-pipelines	Format conversion, preprocessing, and ML dataset preparation

Installing

Copy the skills you want into your project's .claude/skills/ directory or your personal ~/.claude/skills/ directory.

# Install to a project
cp -r dicom-processing /path/to/your/project/.claude/skills/

# Install globally (available in all projects)
cp -r dicom-processing ~/.claude/skills/

# Install all skills globally
for skill in aurabox-rest-api dicom-processing dicomweb-protocol medical-imaging-pipelines; do
  cp -r "$skill" ~/.claude/skills/
done

Restart your agent after copying. Skills trigger automatically based on your prompts.

Skill details

aurabox-rest-api

For developers integrating with the Aurabox platform programmatically.

• Full API reference for patients, cases (de-identified patients), and studies
• Working client code in Python, TypeScript, and curl
• Authentication, pagination, and error handling
• Data models with field types and validation rules

Depends on: An Aurabox account and API key

dicom-processing

For anyone writing code that reads or manipulates DICOM files.

• pydicom fundamentals: reading, writing, modifying DICOM datasets
• Tag reference table with common tags, VRs, and data types
• Pixel data access, windowing, and display
• Transfer syntax handling and decompression
• Bulk operations and metadata extraction
• DCMTK command-line tools

Depends on: pip install pydicom (core), optional numpy, pillow, pylibjpeg

dicomweb-protocol

For developers building HTTP-based imaging integrations.

• QIDO-RS: searching for studies, series, and instances
• WADO-RS: retrieving DICOM instances, metadata, and rendered images
• STOW-RS: uploading DICOM via multipart MIME (the hard part)
• Complete DICOMwebClient class
• DICOM JSON parsing helpers
• Server-specific notes for Aurabox, Orthanc, dcm4chee, Google Cloud Healthcare

Depends on: pip install requests (Python examples)

medical-imaging-pipelines

For technical researchers and ML engineers working with imaging datasets.

• DICOM to NIfTI conversion (SimpleITK and nibabel)
• DICOM to PNG/JPEG with windowing
• DICOM to HDF5 for ML datasets
• CT and MRI intensity normalization
• Resampling, cropping, padding
• Metadata manifest generation
• Complete CT research pipeline example
• PyTorch Dataset integration

Depends on: pip install pydicom numpy SimpleITK nibabel pillow (full set)

How skills work

Each skill is a SKILL.md file with YAML frontmatter (name and description). An AI coding agent reads the descriptions at startup to decide when to load each skill. When a prompt matches a skill's description, the full content is loaded into context.

• Skills have zero cost when not triggered (only the description is loaded)
• Multiple skills can coexist without bloating context
• Skills trigger automatically -- no manual invocation needed