The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes npx playwright to facilitate the installation of browser binaries and the execution of end-to-end test suites within the host environment.
[EXTERNAL_DOWNLOADS]: Through the browser_install tool and associated CLI commands, the skill downloads browser binaries from Microsoft's official Playwright distribution infrastructure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection when processing content from untrusted external websites.
Ingestion points: Data enters the agent's context through browser_navigate, browser_snapshot, and browser_console_messages as described in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructional barriers to prevent the agent from following commands embedded in the web content it processes.
Capability inventory: The skill grants the agent high-impact tools including browser_run_code, browser_evaluate, and browser_file_upload.
Sanitization: There is no evidence of sanitization or validation protocols for the data retrieved from browser interactions before it is analyzed by the agent.
[REMOTE_CODE_EXECUTION]: The browser_run_code and browser_evaluate tools allow for the dynamic execution of JavaScript or Playwright snippets, representing a high-privilege execution capability within the browser session.
[CREDENTIALS_UNSAFE]: The skill documentation includes instructions for managing authentication state via auth-state.json. This file frequently contains sensitive session tokens or cookies, posing a risk of exposure if the file is not handled with appropriate security measures.

playwright-browser