foundation-models-on-device
Pass
Audited by Gen Agent Trust Hub on Mar 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a documentation resource for developers. It describes patterns for on-device AI integration, emphasizing privacy and offline support.
- [PROMPT_INJECTION]: While the skill uses system instructions for role-playing (e.g., 'You are a cooking assistant'), it includes security-positive advice to include safety measures in these instructions to handle dangerous requests.
- [DATA_EXFILTRATION]: The skill explicitly focuses on on-device execution where 'no data leaves the device.' No suspicious network operations or hardcoded credentials were identified in the examples.
- [COMMAND_EXECUTION]: Code snippets use standard Swift and SwiftUI patterns. Tool calling is implemented via defined Swift protocols rather than arbitrary shell execution or unsafe dynamic loading.
- [EXTERNAL_DOWNLOADS]: No external packages or remote scripts are referenced or downloaded. All functionality is presented as part of a local system framework.
Audit Metadata