haiou-claude-personality

Warn

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains multiple instructions to execute platform-specific scripts (deploy.ps1, mac-install.sh, linux-install.sh) and provides commands to modify file permissions (chmod +x). It specifically includes a command to bypass macOS Gatekeeper (xattr -d com.apple.quarantine), which is a technique used to execute unverified software.
  • [PROMPT_INJECTION]: The skill is designed to persistently modify the agent's core behavior and system-level identity by overwriting CLAUDE.md and system-prompt.md. This replaces the standard agent instructions with a custom personality, which is a form of persistent behavioral override.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 8, 2026, 12:33 AM
Security Audit — agent-trust-hub — haiou-claude-personality