The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates explicit defensive instructions to prevent data from external sources (like web search results) from overriding the agent's logic or injecting malicious code into generated artifacts. It mandates escaping source text into DOM text nodes and validating link schemes to prevent Cross-Site Scripting (XSS).
[DATA_EXFILTRATION]: The instructions restrict file operations to a dedicated output folder within the authorized workspace. It specifically prohibits writing to sensitive directories such as node_modules, plugin folders, or hidden directories, reducing the risk of unauthorized file modification or data exposure.
[DYNAMIC_EXECUTION]: While the skill's primary purpose is generating executable web code (HTML/JS), it provides strict safety guidelines for the generated output, such as preferring JSON-serializable data patterns and avoiding unsafe DOM manipulation methods like innerHTML.

lesson-generator