engineering-matchmaking-system
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes player-controlled data such as player IDs and game parameters that enter the system context via Redis storage and retrieval. Ingestion points: Player data is ingested in boilerplate/matchmaker.ts (MatchmakingEngine.enqueue) and boilerplate/lobby-manager.ts (LobbyManager.joinLobby). Boundary markers: Absent. The system relies on structured data but lacks explicit prompt-level delimiters or instructions to ignore embedded content. Capability inventory: The skill is restricted to Redis operations (HSET, ZADD, PUBLISH) and does not have shell execution or general network capabilities. Sanitization: Absent. User-provided identifiers are used directly in Redis keys and published via pub/sub without specific validation or escaping.
- [EXTERNAL_DOWNLOADS]: The installation instructions fetch standard libraries from a well-known service registry. Evidence: bun add bullmq ioredis in SKILL.md fetches packages from the NPM registry.
Audit Metadata