The Agent Skills Directory

[COMMAND_EXECUTION]: The script uses node:child_process.spawn to launch browser executables found on the host system. It also supports disabling the browser sandbox through the CHROME_NO_SANDBOX environment variable, which is a significant security risk in many environments.
[REMOTE_CODE_EXECUTION]: The skill includes an evaluate command that allows for the execution of arbitrary JavaScript strings within the browser context using Playwright's page.evaluate() method. This facilitates dynamic code execution based on potentially untrusted input.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection. Ingestion points: Untrusted content is retrieved from external websites via the snapshot, text, and evaluate commands. Boundary markers: No delimiters or explicit instructions are provided to the agent to distinguish between retrieved web content and its primary instructions. Capability inventory: The browser control functionality includes sensitive actions such as click, type, and navigate, which could be maliciously triggered by instructions embedded in a visited webpage. Sanitization: The skill does not perform any sanitization or filtering on the retrieved page content before presenting it to the agent.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the playwright-core package. This is a well-known library for browser automation and is considered a trusted dependency.

browser