skills/zoom/skills/virtual-agent/ios/Gen Agent Trust Hub

virtual-agent/ios

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents an architecture where web content inside a WKWebView communicates with native iOS code via WKScriptMessageHandler. This creates a standard ingestion point for data from a web environment to enter the application's native context.
  • Ingestion points: Data enters the native context via userContentController:didReceiveScriptMessage: through zoomLiveSDKMessageHandler, commonMessageHandler, and support_handoff handlers defined in examples/js-bridge-patterns.md.
  • Boundary markers: The skill does not explicitly instruct on boundary markers or data validation for the incoming message payloads.
  • Capability inventory: The native application processes messages which can trigger logic for exit flows, common event handling, and support handoffs.
  • Sanitization: The provided Swift examples show raw JSON.stringify data being passed to handlers without explicit sanitization instructions.
  • [DYNAMIC_EXECUTION]: The skill utilizes WKUserScript to inject JavaScript strings directly into the webview's environment at runtime. As described in examples/js-bridge-patterns.md, this is used to initialize the zoomCampaignSdk and establish bridge handlers. While this involves dynamic execution, it is the documented and intended mechanism for WKWebView integration.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 09:55 PM
Security Audit — agent-trust-hub — virtual-agent/ios