The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill involves generating standalone .mjs scripts based on provided templates and executing them using node. It also uses page.evaluate() to run JavaScript within the browser context to retrieve element measurements and computed styles.
[COMMAND_EXECUTION]: The skill utilizes shell commands to manage its environment and execution, including node for script execution, pnpm or npx for package management and browser installation, and pkill to terminate persistent browser processes.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download browser binaries from the official Playwright CDN (cdn.playwright.dev). These downloads originate from a well-known service associated with the Microsoft Playwright project.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it processes content from web pages (via localhost, file://, or remote URLs). Maliciously crafted page content could attempt to influence the agent's behavior through the console logs and error messages that the skill is designed to capture and report.
[CREDENTIALS_UNSAFE]: The skill provides mechanisms for authentication bypass by injecting cookies and local storage state using context.addCookies and context.addInitScript. While this is intended for development testing, it involves handling session identifiers which could be exposed if the generated scripts are not deleted as recommended by the skill's rules.

playwright-verify