engineering-game-backend-architecture
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The server boilerplate in boilerplate/server.ts uses a hardcoded default string 'development-secret-change-me' as a fallback for the JWT_SECRET environment variable.
- [COMMAND_EXECUTION]: The instructions describe running the server code using bun, which involves shell command execution. Evidence: 'bun run boilerplate/server.ts' in SKILL.md.
- [INDIRECT_PROMPT_INJECTION]: The server architecture ingests untrusted client messages via WebSockets, creating a potential vector for indirect prompt injection. 1. Ingestion points: The WebSocket handler in boilerplate/server.ts receiving ClientMessage payloads. 2. Boundary markers: No delimiters or sanitization logic is implemented for message content in the boilerplate. 3. Capability inventory: The skill manages room state and performs network I/O. 4. Sanitization: Includes JWT authentication but lacks application-level input filtering for game actions or chat messages.
Audit Metadata