mobile-push-notifications

Pass

Audited by Gen Agent Trust Hub on Aug 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a surface for indirect prompt injection by processing external data from push notification payloads without extensive validation or sanitization before using it for navigation (deep linking).
  • Ingestion points: The skill processes data from RemoteMessage.data (Flutter/React Native) and UNNotificationResponse.userInfo (iOS) in SKILL.md and references/push-notifications.md.
  • Boundary markers: The implementation lacks boundary markers or explicit instructions for the agent to ignore embedded commands within the notification data.
  • Capability inventory: The skill utilizes navigation capabilities (e.g., navigatorKey.currentState?.pushNamed(deepLink) in Flutter, navigate(remoteMessage.data?.deep_link) in React Native) which could be exploited to redirect the user or agent to sensitive areas if the payload is malicious.
  • Sanitization: The provided snippets perform type checking and null checks but do not include sanitization or verification against a whitelist of allowed deep links.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 23, 2026, 08:00 PM
Security Audit — agent-trust-hub — mobile-push-notifications