The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided brand identity data (name, tagline, tone) and persists it to brand-config.md. This data is later retrieved to generate documents and code. Without explicit boundary markers in the configuration file, there is a potential for instructions embedded in the brand profile to influence future outputs.
Ingestion points: User responses to the conversational setup flow in SKILL.md are saved to brand-config.md.
Boundary markers: The skill lacks explicit instructions for sanitizing or wrapping the user-provided values in brand-config.md.
Capability inventory: The skill writes files to the output directory and triggers generation logic in referenced skills (docx, pptx).
Sanitization: No specific sanitization logic is defined for the user-provided strings.
[DYNAMIC_EXECUTION]: The logo generation process in references/logo-creator.md involves generating SVG code and optionally executing a Python script (using cairosvg or Pillow) to convert the SVG to a PNG file. This is a standard functional requirement for image processing and does not represent a security risk in this context.
Evidence: Python code snippet in references/logo-creator.md demonstrating SVG to PNG conversion.

brand-studio