proteinmpnn-nim
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill transmits protein structure data to NVIDIA's hosted API at health.api.nvidia.com. This is the primary function of the skill and is documented with mandatory user confirmation steps. Authentication is handled using the NGC_API_KEY environment variable, which is sent exclusively via the HTTPS Authorization header to the official vendor endpoint.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external PDB file content and API responses, creating a theoretical attack surface.
- Ingestion points: scripts/design.py reads PDB files from the local filesystem.
- Boundary markers: The skill instructs the agent to obtain user approval before network transmission.
- Capability inventory: Network access via requests and filesystem write access for saving results.
- Sanitization: The script performs basic structural validation of PDB records and amino acid sequences to ensure data integrity.
- [COMMAND_EXECUTION]: Documentation is provided for local Docker deployment and script execution. These instructions are intended for manual user operation to set up the inference environment or run the provided Python client.
Audit Metadata