browser-cdp

Fail

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: HIGHCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is designed to harvest sensitive authentication information and session data.
  • In SKILL.md, the documentation provides specific commands to extract authentication tokens from localStorage and document.cookie using browser evaluation tools.
  • The setup-cdp-chrome.js script performs bulk copying of sensitive Chrome profile data to a new directory. This includes files such as Cookies, Login Data (which contains stored passwords), and Web Data.
  • [COMMAND_EXECUTION]: The setup-cdp-chrome.js script makes extensive use of child_process.execSync to execute system-level commands across different platforms.
  • On Linux and macOS, it executes pgrep, pkill, ps, lsof, ss, and fuser to monitor and terminate processes.
  • On Windows, it utilizes tasklist, taskkill, wmic, and powershell (specifically Get-NetTCPConnection and Get-CimInstance) for similar process and network state management.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the automated processing of arbitrary web content. This creates a significant surface for indirect prompt injection, where malicious instructions hidden on a website could influence the agent's behavior while it is controlling the browser session.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 1, 2026, 09:13 AM
Security Audit — agent-trust-hub — browser-cdp