deepstream-dev
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions and Python code examples follow standard NVIDIA DeepStream development practices for constructing hardware-accelerated video analytics pipelines.
- [EXTERNAL_DOWNLOADS]: The documentation guides the user to fetch pre-trained models from the official NVIDIA NGC registry (
api.ngc.nvidia.com) and adds repositories for well-known services like Confluent for Kafka protocol support. - [COMMAND_EXECUTION]: Includes standard setup procedures such as
sudo apt-get installfor system-level SDK dependencies (e.g.,librdkafka-dev,libmosquitto1) andpip installfor required Python libraries. - [DATA_EXFILTRATION]: Provides implementation patterns for transmitting inference metadata to message brokers (Kafka/MQTT) and external REST endpoints, which is a core functional requirement of the DeepStream SDK for real-time analytics.
- [PROMPT_INJECTION]: No patterns of prompt injection, behavior overrides, or instructions to bypass safety guidelines were detected in the skill content or metadata.
Audit Metadata