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