haiou2-claude-personality
Fail
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill's primary objective is to bypass the agent's safety guidelines and default technical constraints. It provides instructions to eliminate 'cautious' behavior and disclaimers, and includes few-shot examples that demonstrate direct output for potentially harmful requests like exploit development and SQL injection scripts.\n- [COMMAND_EXECUTION]: The installation and uninstallation procedures rely on the execution of multiple scripts (.sh, .ps1, .bat, .command) that modify application configuration files in hidden user directories.\n- [COMMAND_EXECUTION]: The skill explicitly instructs users to disable or bypass system security mechanisms, specifically PowerShell's ExecutionPolicy and macOS's Gatekeeper (quarantine attributes), to enable the execution of its deployment scripts.
Recommendations
- AI detected serious security threats
Audit Metadata