rtvi-cv-scaffold-vss-service

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill generates and utilizes shell commands including docker, docker-compose, and python3 for building and deploying the perception service. A generated utility, tools/kafka_smoketest.py, uses docker exec to perform connectivity checks against Kafka containers.\n- [EXTERNAL_DOWNLOADS]: The skill references the official NVIDIA-AI-Blueprints GitHub organization for blueprint code and the nvcr.io registry for DeepStream base images. These resources are provided by the vendor.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a service that ingests video and Kafka metadata, creating a potential attack surface for indirect prompt injection. However, the agent's role is restricted to infrastructure scaffolding and connectivity testing.\n
  • Ingestion points: The generated tools/kafka_smoketest.py reads data from the mdx-raw Kafka topic.\n
  • Boundary markers: The validation script checks for message presence and length without parsing or executing the payload content.\n
  • Capability inventory: The skill uses docker exec and subprocess.run for service validation.\n
  • Sanitization: The generated scripts include validation for container names and server addresses to prevent command injection via tool arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 05:54 PM
Security Audit — agent-trust-hub — rtvi-cv-scaffold-vss-service