brand-logo
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE]: The skill is configured to read from a file named
DESIGN.md(or a design system file) to extract brand-specific constants like hex color codes, font styles, and proportions. This is a standard functional requirement for brand consistency and does not involve accessing sensitive system credentials or private user data. - [PROMPT_INJECTION]: The skill includes instructions to use 'negative prompts' when interacting with image generation models. These are intended to exclude unwanted visual artifacts (such as 3D effects, gradients, or illegible text) and do not attempt to bypass the safety filters of the underlying language model.
- [COMMAND_EXECUTION]: The workflow involves using built-in image generation tools (e.g.,
image_generate). The inputs to these tools are strictly constrained to design-related keywords and brand tokens, with no evidence of shell command injection or arbitrary code execution. - [EXTERNAL_DOWNLOADS]: The skill does not perform any network operations, remote script downloads, or package installations. All references to external software like Figma, Illustrator, or Canva are for manual human hand-off during the vectorization stage.
Audit Metadata