openfold3-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 ingests outsider-authored free text when it posts user-provided fields like sequence, alignment (MSA msa.*.*.alignment), smiles, ccd_codes, or even structural_templates[*].structure as JSON to the OpenFold3 hosted/local /predict endpoint; the skill then saves/prints parts of the response artifacts returned from that call.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's local startup runs "docker run ... nvcr.io/nim/openfold/openfold3:latest", which will fetch and execute a remote container image (nvcr.io/nim/openfold/openfold3:latest) at runtime.