The Agent Skills Directory

[SAFE]: The skill exhibits an exceptional security architecture with dedicated sanitization and validation logic for every data ingestion point. \n- [INDIRECT_PROMPT_INJECTION]: The potential for indirect injection through untrusted SVG or palette data is thoroughly mitigated by custom parsers in scripts/_svg.py and scripts/_css.py that enforce a strict whitelist of allowed elements and properties, specifically excluding scripts and external resource references. \n- [COMMAND_EXECUTION]: Command execution is restricted to standard project tasks (builds and tests) and is protected by a sophisticated process management utility (scripts/_proc.py) that ensures secure process lifecycle management and cleanup. \n- [DYNAMIC_EXECUTION]: The skill performs dynamic verification of its own edits by importing project files into controlled Node.js or Python execution environments, ensuring system state integrity without introducing arbitrary execution risks.

website-brand-kit