axiom-foundation-models
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a technical guide for developers, providing code patterns for summarization, structured data extraction, and tool calling within the Foundation Models framework.
- [PROMPT_INJECTION]: The instructions include a dedicated security section that warns against prompt injection, specifically instructing developers to never interpolate untrusted user input into the system instructions for an AI session.
- [DATA_EXFILTRATION]: The skill advocates for user privacy by emphasizing the benefits of on-device processing for sensitive data (e.g., medical or financial notes) over sending that data to external server-side APIs.
- [COMMAND_EXECUTION]: No unauthorized or suspicious command executions are present. The provided code snippets utilize standard Swift development patterns and system frameworks such as WeatherKit and CoreLocation.
- [EXTERNAL_DOWNLOADS]: The skill does not request or perform any external downloads of untrusted code or scripts. All dependencies referenced are standard Apple system APIs.
- [INDIRECT_PROMPT_INJECTION]: While the skill describes processing user-provided content, it provides explicit remediation guidance, such as using constrained decoding (@Generable) and prompt/instruction separation to mitigate risks associated with untrusted data ingestion.
Audit Metadata