Skills
Navigate
skills/ffsshhttiikk/opencode-agents-skills/push-notifications/Gen Agent Trust Hub

push-notifications

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [DATA_EXFILTRATION]: The skill examples log sensitive identifiers and notification content to the console, which can lead to data exposure through system logs.
  • Evidence: console.log('Device token:', token) in SKILL.md (Example 1).
  • Evidence: console.log('Storing message:', data) in SKILL.md (Example 2).
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via notification data that influences application flow (Deep Link injection).
  • Ingestion points: messaging().onMessage, messaging().setBackgroundMessageHandler, messaging().getInitialNotification, and messaging().onNotificationOpenedApp in SKILL.md (Example 2).
  • Boundary markers: Absent. The code processes incoming data objects without validation or explicit instructions to ignore embedded commands.
  • Capability inventory: The skill can trigger navigation to arbitrary screens with arbitrary parameters using navigationRef.navigate(screen, params) in SKILL.md (Example 2).
  • Sanitization: Absent. The parseNotificationData method extracts the screen and params directly from the untrusted data object without verification against a whitelist.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 05:19 AM
Security Audit — agent-trust-hub — push-notifications