openfold3-nim

Pass

Audited by Gen Agent Trust Hub on Oct 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests biological sequence data and Multiple Sequence Alignments (MSAs) which are interpolated into API request payloads.
  • Ingestion points: The sequence and msa fields in the inputs payload defined in SKILL.md and references/api.md.
  • Boundary markers: MSAs are required to include a FASTA-style header (>query\n) which serves as a basic delimiter for the alignment data.
  • Capability inventory: The skill utilizes Write permissions to save structure artifacts (PDB/CIF) and Bash for Docker container management and health checks.
  • Sanitization: No explicit sanitization of sequence characters is shown in the provided script examples before payload construction.
  • [COMMAND_EXECUTION]: The skill provides shell commands for environment setup, authentication, and container deployment.
  • Evidence: SKILL.md and references/api.md contain instructions for docker login, docker run, and directory permissions management (chmod 755). These are legitimate operational steps for the local NIM microservice.
  • [EXTERNAL_DOWNLOADS]: The skill pulls a container image from a remote registry during the setup process.
  • Evidence: Instructions include pulling nvcr.io/nim/openfold/openfold3:latest. The nvcr.io domain is the official NVIDIA Container Registry, which is a vendor-owned resource.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 1, 2026, 06:54 AM