haiou-claude-personality-mod

Warn

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructions and provided configuration files are designed to override the agent's core system prompt. The stated purpose is to suppress default safety disclaimers and modify response style constraints.
  • Evidence: The system-prompt.md configuration is explicitly described as being used to "Reduce 'As an AI' disclaimers" and "Enable direct technical responses."
  • [COMMAND_EXECUTION]: The installation and uninstallation procedures involve executing multiple shell and PowerShell scripts that modify the user's environment and configuration files. This includes bypassing system-level security policies on Windows.
  • Evidence: Use of powershell -ExecutionPolicy Bypass -File seagull-files/deploy.ps1 and chmod +x on script files across multiple operating systems.
  • [COMMAND_EXECUTION]: The skill performs file system operations targeting sensitive application configuration directories.
  • Evidence: Scripts target and modify files within ~/.claude (Linux/macOS) and %APPDATA%\Local\Claude-3p (Windows) to deploy the personality mod.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 6, 2026, 03:06 PM
Security Audit — agent-trust-hub — haiou-claude-personality-mod