The Agent Skills Directory

[SAFE]: The skill documents the installation and use of the bx-jsoup module, a standard component for HTML processing within the BoxLang ecosystem. The installation commands (install-bx-module and box install) are legitimate package management operations for this environment.
[SAFE]: The skill focuses on security-enhancing features, specifically providing detailed guidance on using htmlClean() to sanitize untrusted input using predefined safelists (e.g., 'none', 'basic', 'relaxed'). This promotes proper defense-in-depth against XSS attacks.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools to process external HTML data, which is an inherent surface for indirect prompt injection. However, the presence of sanitization utilities and the lack of dangerous capabilities (like network exfiltration or file system modifications) within the skill itself mitigate this risk.
Ingestion points: htmlParse(html) and htmlClean(html) in SKILL.md.
Boundary markers: None explicitly shown for prompt interpolation, but the library acts as a structural boundary itself.
Capability inventory: No command execution, file writing, or unauthorized network operations are defined in the skill.
Sanitization: The htmlClean() function is explicitly featured to remove unsafe tags and attributes from external data.

bx-jsoup