reverse-browser-automation

Warn

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions for the on-demand bootstrap of the OpenReverse framework, which involves cloning an external third-party repository (github.com/zhexulong/openreverse.git) and executing npm install followed by various initialization scripts (npm run init:agents).
  • [EXTERNAL_DOWNLOADS]: Facilitates the installation of the agent-browser package from NPM and the automated download of browser binaries via npx playwright install.
  • [DATA_EXFILTRATION]: Provides specific Node.js code templates designed to extract sensitive information from browser sessions, including session cookies and localStorage contents.
  • [DYNAMIC_EXECUTION]: Encourages the creation and execution of custom Node.js scripts for Playwright that perform advanced interactions such as intercepting network requests, modifying API responses, and injecting payloads into web pages.
  • [COMMAND_EXECUTION]: The skill uses agent-browser and OpenReverse (UIA/CUA modes) to drive interactions with system applications, which allows for the automation of high-privilege tools such as debuggers (x64dbg), disassemblers (IDA Pro), and network analyzers (Wireshark).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 27, 2026, 05:24 AM