skills/thienqo/ios-skills/gamekit/Gen Agent Trust Hub

gamekit

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted match data from remote players through various GameKit delegates and listeners, which represents a potential attack surface for indirect instructions.\n
  • Ingestion points: External data enters the system through delegate methods such as GKMatchDelegate.match(_:didReceive:fromRemotePlayer:), GKLocalPlayerListener.player(_:receivedTurnEventFor:didBecomeActive:), and GKTurnBasedEventListener.player(_:receivedExchangeRequest:for:).\n
  • Boundary markers: The instructions explicitly advise developers to "Treat network and match data as untrusted" and suggest versioning the payload to distinguish between state updates and prevent confusion.\n
  • Capability inventory: The implementation code provided is restricted to GameKit-specific APIs for matchmaking and turn-based data exchange. There are no capabilities for shell command execution, file system modification outside of Game Center saved games, or network operations to untrusted third-party domains.\n
  • Sanitization: The skill advocates for the validation of all match payloads before they are used to mutate game state, following the principle of defensive programming for networked data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 12:01 PM