apple-foundation-models
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains a sanitization function in
references/prompt-design-and-safety.mdthat includes strings typically associated with prompt injection (e.g., "ignore previous instructions"). These strings are used exclusively as examples of defensive sanitization logic and do not represent a malicious attempt to override agent behavior. - [DATA_EXFILTRATION]: A code example in
references/routing-local-vs-bigger-model.mdillustrates routing data to a cloud API (https://your-backend.example.com). This is documented as a standard architectural pattern for hybrid on-device and cloud model usage, accompanied by clear developer guidance on privacy requirements and user notification.
Audit Metadata