apple-on-device-ai
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of technical documentation and code examples for building machine learning features on iOS and macOS. No malicious behavior, obfuscation, or unauthorized data access was detected.
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing user input through AI models, which is an inherent risk surface. However, the documentation correctly identifies this and provides mitigation strategies: 1. Ingestion points: User prompts are passed to LanguageModelSession in references/foundation-models.md. 2. Boundary markers: The skill instructs developers to keep untrusted content in prompts and never in instructions. 3. Capability inventory: Use of model-autonomous tools via the Tool protocol. 4. Sanitization: Use of @Generable schemas and guardrail error handling.
- [EXTERNAL_DOWNLOADS]: The skill references the coremltools library, which is a well-known tool from a trusted organization (Apple).
Audit Metadata