tzai-email-header
Warn
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill implements a search pattern to find an executable script named
tzai-imagein various hidden directories within the user's home folder (e.g.,~/.agents/,~/.claude/,~/.cursor/). Once a match is found, the path is stored in a variable and executed viabash. - [COMMAND_EXECUTION]: User-provided input is interpolated directly into a shell command (
--prompt "<user subject>"). This creates a command injection surface where special characters like backticks or subshell markers in the user's prompt could be executed by the shell if not properly escaped by the agent. - [EXTERNAL_DOWNLOADS]: The skill includes instructions to install an external package from the developer's repository using
npx, which involves fetching and executing code from a remote source.
Audit Metadata