molmim-nim
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill integrates with official NVIDIA BioNeMo NIM services. All API endpoints (health.api.nvidia.com) and container images (nvcr.io/nim/nvidia/molmim) are verified as vendor-owned resources.
- [DATA_EXFILTRATION]: The skill handles authentication for hosted API calls using environment variables. It adheres to secure practices by instructing the agent to load keys from environment variables or .env files rather than hardcoding sensitive information.
- [COMMAND_EXECUTION]: The skill contains shell command templates for Docker operations required to deploy the MolMIM NIM locally. These commands are limited to legitimate infrastructure setup, such as authentication to the NVIDIA Container Registry and starting the container.
- [INDIRECT_PROMPT_INJECTION]: The skill processes molecular SMILES data received from API responses. 1. Ingestion points: API response JSON in SKILL.md and references/api.md. 2. Boundary markers: Absent. 3. Capability inventory: network operations (requests), file writes (Path), and bash commands (docker). 4. Sanitization: Absent (direct parsing of SMILES). This surface presents a low risk of malicious instruction execution.
Audit Metadata