bx-unsafe-evaluate
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documents the
evaluate()Built-in Function in BoxLang, which allows for the dynamic evaluation of strings as code. This capability facilitates dynamic execution, which can lead to code injection vulnerabilities if not handled with extreme caution. - [INDIRECT_PROMPT_INJECTION]: Providing the
evaluate()function creates a risk for indirect prompt injection. If user-controlled data is passed to this function, it could execute arbitrary logic. The skill mitigates this by identifying ingestion points in its documentation and providing a clear evidence chain: - Ingestion points: Expressions containing dynamic or user-supplied data (SKILL.md).
- Boundary markers: The documentation includes 'danger' hints and explicit warnings against using user input.
- Capability inventory: The skill provides dynamic BoxLang execution and shell-based installation commands.
- Sanitization: The skill provides explicit 'Safer Alternatives' (struct access, variables scope, and
invoke()) for the agent to use instead ofevaluate(). - [COMMAND_EXECUTION]: The installation instructions utilize standard CommandBox and BoxLang CLI commands (
install-bx-module,box install). These are documented for environment setup.
Audit Metadata