media-sd

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFEPROMPT_INJECTIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes user-provided text prompts and ComfyUI workflow JSON files as input for image generation. This represents a typical ingestion surface for untrusted data in generative AI applications.
  • [REMOTE_CODE_EXECUTION]: The main driver script sd.py generates and executes localized Python snippets at runtime to manage model weight downloads. This dynamic execution is used to lazily invoke huggingface_hub functionality without requiring it as a mandatory top-level dependency.
  • [COMMAND_EXECUTION]: The skill utilizes subprocess calls to execute system utilities for hardware detection (such as nvidia-smi and uname) and to run internal subcommands for weight management and installation checks.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download of large model assets from HuggingFace and provides instructions for cloning the ComfyUI repository from GitHub. These interactions target well-known and reputable services within the AI development ecosystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 09:54 AM
Security Audit — agent-trust-hub — media-sd