realtime-websocket-patterns
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for real-time communication that process untrusted data from external clients, creating a potential surface for indirect prompt injection.
- Ingestion points: The
websocket_endpointinSKILL.mdusesawait websocket.receive_json()to ingest messages directly from connected clients. - Boundary markers: The provided implementation does not include delimiters or specific instructions to help an agent distinguish between system instructions and data received via WebSockets.
- Capability inventory: The skill includes code for WebSocket connection management, Redis-based pub/sub broadcasting, and SSE event streaming via
fastapiandsse-starlette. - Sanitization: The code snippets demonstrate basic JSON parsing and broadcasting but do not implement sanitization or validation filters to prevent malicious payloads from influencing downstream agent logic.
Audit Metadata