callkit
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 implementation patterns for handling external VoIP push notification payloads, creating a surface for potential indirect prompt injection attacks.
- Ingestion points:
references/callkit-patterns.mdcontains aNotificationServiceextension that decrypts and processes data fromUNNotificationRequestpayloads. - Boundary markers: The provided Swift code lacks delimiters or explicit instructions to treat external metadata (such as caller names) as untrusted content.
- Capability inventory: The skill utilizes
CXProviderto report incoming call information to the system UI based on data received from external push payloads. - Sanitization: No input sanitization or validation logic is demonstrated for the strings derived from the external payloads.
- [EXTERNAL_DOWNLOADS]: The skill includes references to a non-standard external domain for technical documentation.
- Evidence:
SKILL.mdcontains multiple links tohttps://sosumi.ai/documentation/for CallKit and PushKit references.
Audit Metadata