laohan-xiazai
Fail
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to download binary executables (DMG and EXE files) from a non-official GitHub repository (
hanzhcn/laohan-skills).\n- [COMMAND_EXECUTION]: The skill requires several privileged system operations to function:\n - It uses
sudo(Mac) andcertutil(Windows) to install a Root CA certificate (SunnyRoot.cer) into the system's root trust store. This allows the software to decrypt and intercept any HTTPS traffic on the machine.\n - It sets up a local MITM proxy to intercept WeChat network traffic and inject content into the application.\n
- It uses
launchctlto start and manage system services on macOS.\n- [REMOTE_CODE_EXECUTION]: The skill recommends executing global software updates usingnpm update -g @jackwener/opencliandnpx skills update, which downloads and runs unverified code from the internet at runtime.\n- [COMMAND_EXECUTION]: The skill uses a custom wrapper script located at~/.local/bin/视频号下载that executes commands with elevated privileges (sudo).
Recommendations
- AI detected serious security threats
Audit Metadata