azure-architecture-autopilot

Warn

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill extensively uses the powershell tool and the az CLI to perform Azure resource scanning, account management, and infrastructure deployment.
  • [REMOTE_CODE_EXECUTION]: The scripts/cli.py file includes a utility that generates a Node.js script at runtime and executes it via subprocess.run to convert HTML diagrams to PNG format. This dynamic generation uses string interpolation for file paths, creating a risk of code injection if user-provided project or resource names are malicious.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection in its 'Analyze & Modify' path. It reads data from existing Azure environments via az resource list. Maliciously named Azure resources could contain instructions designed to override agent behavior during Bicep generation or architectural advice phases.
  • [EXTERNAL_DOWNLOADS]: The skill fetches documentation and technical specifications from learn.microsoft.com and azure.microsoft.com. These are recognized as trusted official sources for Azure configuration data.
  • [COMMAND_EXECUTION]: The skill's diagram engine relies on executing local python and node binaries to process and render architectural visualizations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 25, 2026, 01:57 AM