datamol

Pass

Audited by Gen Agent Trust Hub on May 28, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill serves as legitimate documentation for the datamol cheminformatics library, providing useful examples for molecular data processing.
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it documents tools for ingesting data from external, potentially untrusted sources like CSV, SDF, and remote URLs. Ingestion points: Functions such as dm.read_sdf, dm.read_csv, and dm.open_df in references/io_module.md. Boundary markers: No explicit delimiters or instructions to ignore embedded natural language commands are identified in the documentation. Capability inventory: The skill allows for file writing (dm.to_sdf, dm.save_df), network operations (via fsspec), and molecular visualization (dm.viz.to_image). Sanitization: While the library performs chemical validation and standardization, it lacks sanitization for potential natural language instructions embedded within data fields.
  • [EXTERNAL_DOWNLOADS]: Documents standard installation procedures for the datamol package and features for fetching datasets from remote protocols (HTTP, S3, GCS) as part of its core functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
May 28, 2026, 03:42 PM
Security Audit — agent-trust-hub — datamol