haiou-claude-code-personality
Fail
Audited by Gen Agent Trust Hub on Jul 8, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides configuration overrides (CLAUDE.md and system-prompt.md) that explicitly instruct the AI to bypass safety guidelines, remove corporate politeness, and eliminate disclaimers or "As an AI" preambles.\n- [PROMPT_INJECTION]: It uses an extensive collection of over 1700 few-shot conversation examples to forcefully anchor a persona that avoids default model constraints and behavioral guardrails.\n- [COMMAND_EXECUTION]: The instructions direct users to execute various platform-specific scripts (e.g., deploy.ps1, mac-install.sh, linux-install.sh) and provide commands to bypass critical OS-level security protections, such as removing the macOS quarantine flag (xattr -d com.apple.quarantine) and lowering PowerShell execution policies (Set-ExecutionPolicy -ExecutionPolicy RemoteSigned).\n- [COMMAND_EXECUTION]: The skill includes instructions for the forceful termination of system processes using pkill and taskkill to restart the host application, which can be disruptive to the local environment.
Recommendations
- AI detected serious security threats
Audit Metadata