deepstream-dev

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download software dependencies and pre-trained AI models from official NVIDIA and well-known providers like Confluent, including NVIDIA's NGC registry.
  • [PROMPT_INJECTION]: The skill architecture is designed to ingest and process untrusted data from video sources and RTSP streams, representing an indirect prompt injection surface.
  • Ingestion points: nvurisrcbin and nvmultiurisrcbin components described in SKILL.md and references/streaming_sources.md.
  • Boundary markers: None explicitly implemented in the provided code examples to separate control instructions from stream data.
  • Capability inventory: The pipelines support network operations via nvmsgbroker and requests, and file output via filesink as documented in references/kafka_messaging.md and references/buffer_apis.md.
  • Sanitization: Provided code patterns do not include explicit validation or sanitization of input URIs or stream metadata.
  • [COMMAND_EXECUTION]: Development and environment setup instructions include shell commands for repository configuration, key management, and package installation using sudo privileges.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 01:40 AM
Security Audit — agent-trust-hub — deepstream-dev