geniml

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches the geniml package and its development version from the official GitHub repository (github.com/databio/geniml).
  • [EXTERNAL_DOWNLOADS]: References the StarSpace tool from Facebook Research's public repository for embedding tasks.
  • [EXTERNAL_DOWNLOADS]: Downloads pre-trained models from the official Hugging Face repository (databio/scembed-pbmc-10k).
  • [PROMPT_INJECTION]: The skill processes external genomic data (BED files) and metadata CSVs, creating a surface for indirect prompt injection where malicious content in the data could attempt to influence the agent's behavior.
  • Ingestion points: SKILL.md (metadata.csv, bed_files/), references/scembed.md (scatac_data.h5ad)
  • Boundary markers: Absent for data interpolation.
  • Capability inventory: File system access for reading/writing results, command-line execution of genomic tools.
  • Sanitization: Data is processed through machine learning tokenization and embedding models.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 11:42 AM
Security Audit — agent-trust-hub — geniml