foundation-models-on-device
Pass
Audited by Gen Agent Trust Hub on Mar 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely instructional documentation for a native Swift framework (FoundationModels). No executable code, scripts, or automated tools are shipped with the skill.
- [DATA_EXFILTRATION]: All code patterns focus on on-device execution to ensure privacy. No network requests, hardcoded credentials, or exfiltration patterns were detected.
- [PROMPT_INJECTION]: The provided prompts and instructions are benign, serving to define model roles (e.g., 'cooking assistant') and enforce safety measures rather than bypassing them.
- [REMOTE_CODE_EXECUTION]: No remote script downloads, piped executions, or unverifiable external dependencies are referenced. All examples use standard Swift syntax and native framework calls.
- [COMMAND_EXECUTION]: The skill does not contain any shell commands, privilege escalation attempts, or persistence mechanisms.
Audit Metadata