The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the bx-esapi module via install-bx-module and box install. These are standard package management commands for the BoxLang and CommandBox ecosystem. The module being installed is a core security component provided by the skill author's organization.
[INDIRECT_PROMPT_INJECTION]: The skill documentation describes techniques for handling untrusted data, creating an attack surface for indirect prompt injection if improperly implemented by an agent.
Ingestion points: The skill uses examples involving user-controlled input such as user.displayName, user.bio, user.website, and userHtmlInput within SKILL.md.
Boundary markers: The skill explicitly instructs on the use of boundary markers through context-aware encoding functions (e.g., encodeForHTML, encodeForJavaScript).
Capability inventory: The skill itself does not implement dangerous capabilities like arbitrary shell execution, file system writes, or unauthorized network operations.
Sanitization: The primary purpose of the skill is to provide sanitization and encoding mechanisms (e.g., AntiSamy via htmlSanitize) to mitigate injection risks. The instructions follow security best practices.

bx-esapi