mobile-camera-vision-camera

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data through QR and barcode scanning functionality.
  • Ingestion points: The onCodeScanned callback in ScannerScreen (examples/scanning-and-processing.md) receives data from external physical sources.
  • Boundary markers: No delimiters or instructions are provided to distinguish between scanned data and system instructions.
  • Capability inventory: The skill can trigger camera captures, access temporary file system paths for media, and modify application state.
  • Sanitization: While debouncing is used for performance, no validation or escaping of the scanned string values is performed before potential use by the agent.
  • [DYNAMIC_EXECUTION]: The skill utilizes the worklet pattern for real-time frame analysis, which involves executing JavaScript on a parallel thread via JSI.
  • Evidence: Multiple files including SKILL.md and examples/scanning-and-processing.md use the mandatory 'worklet' directive in useFrameProcessor and useSkiaFrameProcessor.
  • Risk: While this is a standard and required pattern for the VisionCamera library's performance, it represents a complex runtime execution model.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 01:04 PM