openfold2-nim
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external biological data, including protein sequences, A3M alignments, and mmCIF templates. These represent untrusted ingestion points that could be used for indirect prompt injection, although the specialized formats provide natural constraints.
- Ingestion points: The sequence, alignments, and explicit_templates fields in SKILL.md and references/api.md.
- Boundary markers: No explicit delimiter-based boundary markers are defined in the instructions for these data fields.
- Capability inventory: The skill utilizes the Bash tool for container orchestration and shell-based validation, as well as the requests library for API communication.
- Sanitization: references/validation.md includes guidelines for validating amino-acid symbols and template structure records.
- [COMMAND_EXECUTION]: The skill provides shell commands for local environment setup, including Docker authentication and container deployment.
- Evidence: Scripts in references/api.md demonstrate docker login and docker run usage for NIM microservices.
- [EXTERNAL_DOWNLOADS]: Downloads the OpenFold2 NIM container image from NVIDIA's official registry.
- Evidence: The image nvcr.io/nim/openfold/openfold2:latest is referenced for local deployment in references/api.md.
Audit Metadata