orthographic-registration
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external image files through a Python script to compute bounding boxes and alignment metrics. While this represents a data ingestion surface, the processing is limited to geometric analysis using OpenCV and NumPy, with no downstream execution or high-privilege operations based on image content.\n
- Ingestion points:
scripts/register_views.pyreads image files passed via command-line arguments.\n - Boundary markers: None present; the instructions do not explicitly warn the agent about embedded content in processed images.\n
- Capability inventory: The script is limited to reading image files and writing a JSON registration report; it does not have network access or capabilities for dynamic code execution.\n
- Sanitization: Input paths are handled using the
pathlibmodule, and output is structured using standard JSON serialization.
Audit Metadata