image-brand-compliance

Warn

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill incorporates a reporting mechanism that encodes the content of reviewed files into base64 data URIs within a generated HTML dashboard. This process could be used to expose the contents of sensitive local files if they are improperly passed to the skill as target 'images'.
  • [COMMAND_EXECUTION]: The skill implements an 'Apply Fixes' workflow that generates Python scripts at runtime to perform image editing. These scripts are executed via a shell environment. This dynamic execution pattern poses a risk if the parameters for the generated code are influenced by untrusted external inputs.
  • [EXTERNAL_DOWNLOADS]: The skill requires the 'Pillow' library for image manipulation and directs users to install it via standard package management if it is not found on the system.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted external data. Ingestion points: The skill reads image files and brand guideline documents provided as file paths or text. Boundary markers: No explicit delimiters are used to separate input data from instructions. Capability inventory: The skill can read local files and execute dynamic Python code via subprocesses. Sanitization: There are no documented validation or escaping steps for data extracted from images or guidelines before its use in analysis or script generation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 6, 2026, 05:59 AM
Security Audit — agent-trust-hub — image-brand-compliance