homekit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill includes multiple reference links to documentation on the domain
sosumi.ai. These links are used to provide additional context for framework-specific classes and methods. Although this domain is not a recognized official source, its usage here is strictly informational and does not involve automated code execution. - [INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing data from external smart-home accessories, which could serve as an ingestion point for untrusted input.
- Ingestion points: Accessory metadata, room names, and characteristic values are retrieved from the
HMHomeandHMAccessoryobjects as shown in SKILL.md and references/matter-commissioning.md. - Boundary markers: The provided Swift code snippets do not implement delimiters or safety instructions to prevent the agent from interpreting potentially malicious strings found in accessory characteristics.
- Capability inventory: The skill includes templates for network requests (
fetchRoomsFromBackend,saveDeviceToBackend), accessory control (writeValue), and UI rendering. - Sanitization: There is no explicit sanitization or validation logic for external strings before they are processed or displayed in the application context.
Audit Metadata