proteinmpnn-nim

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow sends the staged /workspace/input/1R42.pdb contents inline as the input_pdb JSON field to the ProteinMPNN NIM (local or hosted) without any LLM selecting a specific trusted item first, so an outsider who can influence that file content can inject poison text into the model’s input.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's Local Docker startup explicitly pulls and runs a remote container image (executes remote code) from nvcr.io/nim/ipd/proteinmpnn:latest as part of normal runtime/setup, so this external image URL is a runtime dependency that executes remote code.