foundation-models-on-device
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a documentation guide for developers using Apple's on-device language models. It emphasizes privacy and offline support.
- [PROMPT_INJECTION]: The instructions provide guidance on using system instructions to define model roles and safety constraints. While the
respond(to:)method accepts arbitrary user input (creating an indirect prompt injection surface), this is an inherent property of the described framework rather than a malicious instruction within the skill. - [COMMAND_EXECUTION]: The skill describes a 'Tool Calling' pattern that allows models to invoke local Swift code. The use of the
@Generablemacro for arguments provides compile-time safety and structural validation, which is a security best practice for handling LLM-generated tool arguments.
Audit Metadata