on-device-ai

Pass

Audited by Gen Agent Trust Hub on Nov 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a legitimate technical resource for developers. It utilizes standard React Native and Expo patterns to implement on-device AI features. All instructions and code snippets are consistent with the documented purpose of the library.
  • [EXTERNAL_DOWNLOADS]: The skill mentions downloading machine learning models (.pte files) from remote sources. These operations are handled by the library's resource-fetcher adapters, which target the author's official domains (swmansion.com) or HuggingFace. These are trusted or well-known services for model distribution.
  • [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risks of processing untrusted data by providing the usePrivacyFilter utility, which is designed to detect and redact PII on-device before it can be processed by other components. It also demonstrates how to use structured output validation (Zod) to ensure that AI-generated content conforms to expected schemas, providing a layer of defense against unexpected model behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 2, 2026, 11:52 PM