family-controls

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a pattern for ingesting application and web domain selections through the FamilyActivityPicker. This data is used to configure device-wide restrictions and monitoring schedules. Although this constitutes an ingestion of untrusted user input that influences system behavior, the risk is mitigated by the iOS framework's use of opaque tokens, which prevent the application from accessing raw identifiers or URLs.
  • Ingestion points: FamilyActivitySelection in SKILL.md (AppRestrictionPickerView) and references/screen-time-patterns.md (MyDeviceActivityMonitor).
  • Boundary markers: Not applicable; the skill relies on the system-provided modal UI for input selection.
  • Capability inventory: Modifies system shield settings via ManagedSettingsStore and schedules usage monitoring via DeviceActivityCenter.
  • Sanitization: The implementation utilizes system-enforced ApplicationToken and WebDomainToken objects which are opaque by design, preventing direct inspection or manipulation of the underlying data.
  • [EXTERNAL_DOWNLOADS]: The skill provides informational links to documentation hosted at sosumi.ai. These references are for developer guidance and do not involve automated fetching or execution of remote content within the agent's environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 11:13 AM
Security Audit — agent-trust-hub — family-controls