haiou-seagull-claude-personality
Fail
Audited by Gen Agent Trust Hub on Jul 8, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs users to bypass macOS Gatekeeper security controls by manually removing the quarantine attribute from its scripts using
xattr -d com.apple.quarantine 启动.command. - [COMMAND_EXECUTION]: For Windows users, the skill encourages bypassing PowerShell script execution policies with the
powershell -ExecutionPolicy Bypasscommand, which overrides local security settings designed to prevent malicious script execution. - [PROMPT_INJECTION]: The skill targets the persistent modification of the Claude Code agent's system prompt by overwriting
~/.claude/system-prompt.md. By injecting 1,700+ conversation examples at the system level, the skill can persistently alter the agent's behavior and potentially override its core safety or operational constraints. - [COMMAND_EXECUTION]: The installation process involves running multiple scripts (
deploy.ps1,mac-install.sh,linux-install.sh) that perform high-risk operations including modifying files in protected or sensitive user configuration directories such as~/.claude/and%APPDATA%.
Recommendations
- AI detected serious security threats
Audit Metadata