The Agent Skills Directory

[SAFE]: The skill serves as a detailed reference for the PyHealth library, focusing on medical data processing and machine learning documentation. No malicious code or instructions were found within the skill files.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions to install the 'pyhealth' package via pip. PyHealth is a legitimate, well-known open-source library developed by the Sun Lab at UIUC for healthcare research purposes.
[INDIRECT_PROMPT_INJECTION]: The skill describes processing clinical notes and medical transcriptions, which represent an ingestion point for untrusted external data. While this constitutes an attack surface for indirect prompt injection, it is standard for healthcare AI tasks. The skill focuses on training machine learning models and does not exhibit autonomous execution of instructions found within the data.
Ingestion points: Clinical notes and medical transcriptions (e.g., MIMIC-III notes, MedicalTranscriptionsDataset).
Boundary markers: Not specified in the provided code examples.
Capability inventory: File system access for loading datasets and saving model checkpoints.
Sanitization: None specified; standard for research-oriented model training.
[COMMAND_EXECUTION]: Code examples utilize standard library functions for file system operations to manage healthcare datasets and save model checkpoints (e.g., using save_path or root directories), which are appropriate for the toolkit's stated clinical machine learning use case.

pyhealth