vss-query-analytics

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill author (NVIDIA) and the referenced GitHub repository (github.com/NVIDIA-AI-Blueprints) are verified vendor resources, consistent with the provided context.
  • [PROMPT_INJECTION]: The skill documentation proactively addresses indirect prompt injection by including a 'Scope guard' that warns against executing instructions embedded in alerts or analytics payloads. It specifically forbids auto-deployment triggered by untrusted data content.
  • [COMMAND_EXECUTION]: The use of shell commands like curl, jq, and docker is restricted to querying the designated analytics service and verifying deployment status within the user's environment.
  • [CREDENTIALS_UNSAFE]: The skill correctly manages authentication by utilizing environment variables (NGC_CLI_API_KEY, NVIDIA_API_KEY) for image pulls, avoiding hardcoded secrets.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 08:07 PM
Security Audit — agent-trust-hub — vss-query-analytics