The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documents the evaluate() function, which is a primitive for executing arbitrary strings as code at runtime. The documentation identifies this as a security risk and includes a 'danger' hint warning that it can create injection vulnerabilities if user-controlled input reaches it. It provides safer alternatives like struct key access and the invoke() function, which reduces the risk of an agent defaulting to unsafe execution patterns.
[COMMAND_EXECUTION]: The documentation includes installation instructions for the bx-unsafe-evaluate module using the install-bx-module and box install commands. These are standard package management operations for the BoxLang ecosystem and are appropriate for the skill's role in documenting a platform-specific feature.

bx-unsafe-evaluate