virtual-agent/android

Warn

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill provides Kotlin code examples in examples/js-bridge-patterns.md that utilize webView.loadUrl("javascript:...") to inject and execute dynamic JavaScript strings at runtime within the WebView context.- [INDIRECT_PROMPT_INJECTION]: The skill documents an architecture for ingesting data from a web environment into the native Android application via bridge callbacks such as handleCommon and handleHandoff. This constitutes a potential attack surface where data processed by the agent could contain instructions designed to influence native behavior. Ingestion points: JavaScript bridge listeners defined in examples/js-bridge-patterns.md. Boundary markers: None provided in the implementation examples. Capability inventory: Native method execution including handleExit, handleCommon, and handleHandoff. Sanitization: The examples demonstrate data serialization using JSON.stringify but do not show validation or filtering of the payload content before native processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 26, 2026, 09:55 PM
Security Audit — agent-trust-hub — virtual-agent/android