The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes text data from element attributes or function arguments and renders it using innerHTML without sanitization, creating a surface for indirect injection via malicious HTML or scripts.
Ingestion points: The scrambleTo function in text-scramble.js accepts text via the newText parameter and the data-text attribute.
Boundary markers: No boundary markers are used to delimit the text content or warn the agent about embedded instructions.
Capability inventory: The script uses el.innerHTML to update the DOM, which can execute embedded script tags or render malicious HTML if the input is attacker-controlled.
Sanitization: The implementation lacks any sanitization or escaping logic for the characters being resolved into the final string.

text-scramble