The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses strong instructional framing (e.g., "CORE DIRECTIVE", "IRON RULE", "FORBIDDEN FROM") to constrain the agent's creative output and prevent default behaviors. These are stylistic constraints meant to improve output quality and do not attempt to bypass safety filters or reveal system prompts.
[EXTERNAL_DOWNLOADS]: The instructions reference https://picsum.photos as a source for placeholder images. This is a well-known and standard service for design prototyping and does not represent a security risk.
[DYNAMIC_EXECUTION]: The skill instructs the agent to "simulate a Python script execution" to generate a deterministic seed for layout randomization. This is a reasoning technique performed by the LLM to guide its internal choices and does not involve actual code execution on the underlying operating system or environment.
[COMMAND_EXECUTION]: All mentioned tools and libraries, such as GSAP and Phosphor Icons, are standard frontend development packages intended for UI rendering. There are no attempts to execute system commands or access unauthorized local resources.

gpt-taste