bx-mail
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent on how to send emails, which creates a surface for indirect prompt injection where malicious instructions could be embedded in data processed by the agent.
- Ingestion points: Variables interpolated into
bx:mailattributes (to, from, subject) and the email body content inSKILL.md. - Boundary markers: None provided in the examples to delimit untrusted content from email headers or message structure.
- Capability inventory: Network communication for sending emails via SMTP (
bx:mail) and local file system access for attachments (bx:mailparam). - Sanitization: None described; the skill relies on the agent's safety filters or developer-level validation.
- [EXTERNAL_DOWNLOADS]: Fetches the
bx-mailmodule using standard BoxLang and CommandBox installation commands. This is a legitimate vendor resource setup procedure. - [COMMAND_EXECUTION]: Provides instructions for executing package management commands (
install-bx-module,box install) for the initial setup of the email module.
Audit Metadata