The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The run.js file implements a 'Universal Executor' pattern that accepts arbitrary Node.js code via command-line arguments, files, or standard input. This code is then wrapped in a template and executed using require() on a dynamically generated file, allowing for unrestricted code execution on the host system.
[COMMAND_EXECUTION]: The run.js script uses child_process.execSync to run shell commands for environment setup, including npm install and npx playwright install, which are triggered automatically if the library is not detected.
[EXTERNAL_DOWNLOADS]: The skill downloads the Playwright package and browser binaries (Chromium) from remote npm and Playwright registries during the setup process.
[DATA_EXFILTRATION]: The detectDevServers function in lib/helpers.js performs automated probing of local network ports (such as 3000, 3001, 8080, etc.) to discover running services. This can be used for local environment reconnaissance and identifying internal targets.
[PROMPT_INJECTION]: The skill generates and executes scripts based on user-provided descriptions without implementing boundary markers or sanitization logic. This creates a vulnerability surface where the agent could be manipulated into using the browser to access sensitive local files via file:/// protocols or internal network resources.

playwright-test-automation