skills/bria-ai/bria-skill/image-utils/Gen Agent Trust Hub

image-utils

Pass

Audited by Gen Agent Trust Hub on Sep 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill contains functionality to download image data from external URLs via the ImageUtils.load_from_url method, which utilizes the requests library. This is a standard feature for image processing utilities to handle remote assets.
  • [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for processing untrusted image data from external sources. While it does not process natural language instructions within the images, it possesses the capabilities required for data ingestion.
  • Ingestion points: ImageUtils.load and ImageUtils.load_from_url in references/code-examples/image_utils.py ingest data from URLs, file paths, and base64 strings.
  • Boundary markers: None identified; the skill treats all input as image data for processing.
  • Capability inventory: The skill has the capability to perform network GET requests (requests.get), create local directories (Path.mkdir), and write files to the local file system (Image.save).
  • Sanitization: Standard Pillow Image.open validation is used, which ensures data adheres to supported image formats.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 30, 2026, 05:04 PM