skills/nvidia/skills/diffdock-nim/Gen Agent Trust Hub

diffdock-nim

Pass

Audited by Gen Agent Trust Hub on Oct 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides documented shell commands for local Docker deployment, including docker login and docker run configurations. These commands are intended for environment setup and are directed at official NVIDIA registries (nvcr.io).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external file content (PDB and SDF files) to prepare inputs for the DiffDock API. While this presents a surface for indirect prompt injection, the risk is minimized by the structured nature of the data and the specific filtering for 'ATOM' records.
  • Ingestion points: Reads protein structures from protein.pdb and ligand data from ligand.sdf or SMILES strings.
  • Boundary markers: The skill instructs the agent to filter PDB files for 'ATOM' records only, providing a level of structural validation.
  • Capability inventory: Uses the requests library for API communication and Bash for health checks and Docker management.
  • Sanitization: Implements basic filtering of PDB files to include only 'ATOM' lines before transmission to the API.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 3, 2026, 06:50 PM