The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's delivery instructions in references/delivery.md direct the agent to write and execute scripts for exporting brand assets, such as converting SVG files or generating platform-specific icons. Generating and running code at runtime is a high-capability function that could be manipulated to perform unintended actions.
[INDIRECT_PROMPT_INJECTION]: The skill has a broad attack surface, as it is instructed to process user-provided descriptions, images, and information retrieved from external URLs during the research phase. It lacks clear boundary markers or sanitization for this external data, meaning instructions embedded in brand specifications or case studies could potentially influence the agent's logic or the scripts it generates.
[COMMAND_EXECUTION]: The workflow for asset delivery involves directory creation and the use of system tools or custom scripts to finalize files. These execution capabilities, while functional, provide a pathway for command-related exploits if the underlying data or generated scripts are compromised.

logo-design