anomalib-tiled-ensemble

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external image datasets for training and evaluation, which creates a potential surface for indirect injection via data content or directory structures.
  • Ingestion points: Dataset directory path in SKILL.md under data.init_args.root.
  • Boundary markers: Absent.
  • Capability inventory: Local subprocess execution of training and evaluation scripts (tools/tiled_ensemble/train.py, tools/tiled_ensemble/eval.py).
  • Sanitization: Relies on default library behavior for image loading and processing.
  • [DYNAMIC_EXECUTION]: The configuration utilizes class_path to dynamically load and instantiate Python classes for data modules and model architectures at runtime.
  • Evidence: data.class_path and TrainModels.model.class_path parameters in the YAML configuration example in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 01:43 PM
Security Audit — agent-trust-hub — anomalib-tiled-ensemble