tzai-avatar
Fail
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The "Run" section contains a shell command template that interpolates user-provided text ("") into a bash command string. While the placeholder is wrapped in double quotes, the command remains susceptible to injection attacks via shell expansion (e.g., using backticks or $(...) syntax), which could allow a malicious user to execute arbitrary commands with the agent's privileges.
- [EXTERNAL_DOWNLOADS]: The skill searches for an external executable engine in several hidden directories within the user's home folder. If the engine is not located, it prompts the user to install a specific package ("kedoupi/tzai-image-skill") using npx. This creates a dependency on external code and introduces risks associated with unverified third-party software.
Recommendations
- AI detected serious security threats
Audit Metadata