tzai-badge
Warn
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill presents an indirect prompt injection vulnerability where user-provided text is directly interpolated into a bash command via the
--prompt "<user subject>"argument. This creates a surface for command or argument injection if the agent does not properly escape shell metacharacters. - [COMMAND_EXECUTION]: The skill employs dynamic execution by searching for and executing a script (
$ENGINE) from computed paths based on various platform-specific hidden directories in the user's home folder (e.g.,$HOME/.agents/,$HOME/.cursor/). - [EXTERNAL_DOWNLOADS]: The skill specifies a dependency on the
tzai-imageengine and, if missing, instructs the user to download and install it usingnpx skills add kedoupi/tzai-image-skill, which involves executing code from a remote NPM registry.
Audit Metadata