anomalib-adding-a-datamodule

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill provides templates for creating datamodules that process external file system data. This presents an ingestion surface where an agent interacting with a compromised or malicious directory structure could be influenced by specifically crafted file names or paths.
  • Ingestion points: Data source paths provided to Folder datamodule (normal_dir, abnormal_dir, mask_dir) and root path in MyDataModule.
  • Boundary markers: None implemented in the provided skeleton code.
  • Capability inventory: File system access and metadata parsing via pandas and torch within the AnomalibDataset implementation.
  • Sanitization: No input sanitization or path validation is demonstrated in the implementation guides.
  • [DYNAMIC_EXECUTION]: The skill's testing guide utilizes dynamic execution patterns to dispatch calls to dummy dataset generators.
  • Pattern: The DummyDatasetGenerator.generate_dataset() method uses getattr to invoke format-specific methods based on a string-based data_format value. This introduces a dynamic execution path controlled by the format identifier.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 01:43 PM
Security Audit — agent-trust-hub — anomalib-adding-a-datamodule