weatherkit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for fetching and displaying data from the WeatherKit API, including weather alerts and summaries. While weather data is generally structured, the text-based summaries and alert details provided by external meteorological services represent a potential surface for indirect prompt injection if the agent interprets these strings as instructions during runtime processing.
- Ingestion Points: Data enters the agent's context through
WeatherService.shared.weather(for: location)calls documented inSKILL.mdandreferences/weatherkit-patterns.md. - Boundary Markers: No explicit delimiters or instructions are provided to the agent to treat weather strings as untrusted data; however, the skill relies on standard SwiftUI
TextandLinkcomponents for display. - Capability Inventory: The skill facilitates network communication with Apple's WeatherKit service. It does not contain capabilities for arbitrary file system modification, command execution, or unauthorized network operations.
- Sanitization: There is no evidence of sanitization or filtering of the
alert.summaryorcurrent.condition.descriptionstrings before they are displayed to the user or processed by the model. - [EXTERNAL_DOWNLOADS]: The skill contains multiple references to documentation and video resources hosted on the
sosumi.aidomain (e.g.,https://sosumi.ai/documentation/weatherkit). This domain is not an official Apple property. While it appears to function as a mirror for official documentation, users should be aware that these are third-party links.
Audit Metadata