inference-builder

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [SAFE]: The skill enables standard Vision AI developer workflows, such as configuring and building inference pipelines on NVIDIA hardware. No obfuscation, privilege escalation, or persistence mechanisms were detected.
  • [COMMAND_EXECUTION]: The instruction set includes shell commands for local virtual environment management, dependency installation via pip, and executing a Python-based generator script. These actions are restricted to the local repository context and are necessary for the tool's core functionality.
  • [EXTERNAL_DOWNLOADS]: The skill references downloading pre-trained models from NVIDIA NGC and HuggingFace. These are trusted and well-known repositories for AI assets and do not represent a security risk when used for their intended purpose.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests YAML configurations and OpenAPI specifications as part of the pipeline generation process. While this creates an ingestion surface for external data, it is a primary function of the builder tool and occurs within a developer-controlled local environment.
  • [DYNAMIC_EXECUTION]: The builder utilizes Jinja2 templates to produce application source code and Dockerfiles based on user-supplied configurations, which is a standard and appropriate implementation for code generation tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 09:41 PM