implementing-realtime-sync

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The example backend script examples/llm-streaming-sse/backend.py takes user input via the prompt query parameter and directly interpolates it into requests sent to LLM providers (OpenAI and Anthropic). \n
  • Ingestion points: prompt parameter in the stream_chat route of backend.py.\n
  • Boundary markers: None present; the input is passed directly to the model as a user message.\n
  • Capability inventory: The script performs network operations to external AI services.\n
  • Sanitization: No input validation or escaping is applied to the prompt before it is sent to the API.\n- [EXTERNAL_DOWNLOADS]: The skill's installation instructions and examples involve downloading and installing several third-party libraries from official package registries. \n
  • Evidence: The requirements.txt file in the SSE example lists fastapi, uvicorn, openai, anthropic, and python-dotenv. These are standard, well-known libraries.\n- [COMMAND_EXECUTION]: The example backend in examples/llm-streaming-sse/backend.py configures CORS with allow_origins=[\"*\"]. While the code includes a comment to configure this for production, the default behavior in the provided example allows requests from any origin, which is a common security misconfiguration if deployed as-is.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 12:52 PM
Security Audit — agent-trust-hub — implementing-realtime-sync