tuzi-danger-gemini-web

Warn

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: Launches a web browser (e.g., Chrome, Edge) with the --remote-debugging-port flag enabled to automate the extraction of session cookies through the Chrome DevTools Protocol in load-browser-cookies.ts.
  • [COMMAND_EXECUTION]: Executes system commands (cmd.exe and wslpath) using execSync within paths.ts to resolve directory structures when running in a Windows Subsystem for Linux (WSL) environment.
  • [CREDENTIALS_UNSAFE]: Extracts Google session cookies (__Secure-1PSID, __Secure-1PSIDTS) and persists them in a local JSON file (cookies.json) within the user's application data directory for session reuse.
  • [REMOTE_CODE_EXECUTION]: The execution instructions in SKILL.md suggest using npx -y bun, which automatically downloads and executes the Bun runtime from the npm registry if it is not already present on the system.
  • [DATA_EXFILTRATION]: Performs network requests to Google services (gemini.google.com, googleapis.com) to transmit prompts and upload user-provided files for processing, which is consistent with the skill's primary function of providing AI generation via Google's web platform.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 6, 2026, 03:45 PM