baoyu-post-to-x

Warn

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill frequently executes system-level commands through osascript on macOS, powershell.exe on Windows, and various utilities like xdotool or ydotool on Linux to simulate user keystrokes and activate application windows.
  • [DATA_EXFILTRATION]: The skill uses the Chrome DevTools Protocol (Network.getCookies) to programmatically extract sensitive session cookies, including auth_token and ct0, from the browser's memory to verify the user's X login status.
  • [REMOTE_CODE_EXECUTION]: The skill performs dynamic code generation and execution by writing Swift and PowerShell scripts to temporary files at runtime and then executing them with their respective system interpreters.
  • [EXTERNAL_DOWNLOADS]: The automation logic depends on non-standard npm packages, such as baoyu-chrome-cdp and baoyu-md, which are maintained by the skill's vendor and provide low-level browser control and Markdown processing capabilities.
  • [PROMPT_INJECTION]: The instructions in SKILL.md advise the agent to automatically terminate Chrome processes using pkill without requesting user permission during troubleshooting, which bypasses standard autonomy and confirmation guardrails.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 29, 2026, 05:09 PM
Security Audit — agent-trust-hub — baoyu-post-to-x