playwright

Warn

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONNO_CODEEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides an eval command allowing execution of arbitrary JavaScript within the browser context, presenting a risk of dynamic code execution.
  • [COMMAND_EXECUTION]: Local Bash commands are used to manage a Node.js server process and browsers using pkill.
  • [DATA_EXFILTRATION]: Documentation demonstrates the extraction of sensitive information, such as authentication tokens from localStorage, using browser automation tools.
  • [NO_CODE]: The implementation file playwright-server.js is not provided, which is the primary logic for the browser-to-agent interface.
  • [EXTERNAL_DOWNLOADS]: Fetches Playwright and browser binaries from well-known official sources during the setup process.
  • [PROMPT_INJECTION]: The skill faces indirect prompt injection risks. It ingests untrusted data from web pages (via navigate and snapshot in SKILL.md) without boundary markers or instructions to ignore embedded commands. These inputs could exploit high-privilege tools like eval and click found in references/commands.md, with no sanitization logic present.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 28, 2026, 11:37 AM