The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes text content from the DOM to apply animation wrappers. Although this ingests potentially untrusted data from the page, the implementation is safe.
Ingestion points: The script reads text via element.textContent in the splitWordReveal function in SKILL.md.
Boundary markers: None present.
Capability inventory: The code is limited to client-side DOM manipulation (createElement, appendChild) and class list modification. It lacks network, file system, or system command capabilities.
Sanitization: The script uses textContent and createTextNode to handle all text content, ensuring that any characters or tags in the source text are treated as literal strings and not executable HTML or scripts.
[SAFE]: The implementation adheres to security best practices for client-side utilities. It avoids dangerous methods such as innerHTML or eval(), respects the prefers-reduced-motion media query, and ensures accessibility by using aria-label for screen readers. No signs of obfuscation, unauthorized data exposure, or remote code execution were detected.

staggered-word-reveal