The Agent Skills Directory

[Command Execution and External Downloads]: The skill utilizes npm and npx to manage the installation and execution of the @playwright/cli package. These are standard tools for JavaScript development, and their use here targets official packages from a recognized vendor to enable browser automation.
[Browser State Management]: The skill provides commands such as state-save, cookie-get, and localstorage-list to manage authentication state. While necessary for testing and automating logged-in sessions, users should ensure that the resulting state files (like auth.json) are handled securely and excluded from version control as suggested in the documentation.
[Dynamic Code Execution]: The run-code and eval commands allow for the execution of arbitrary JavaScript within the browser context. This is a powerful feature intended for advanced automation scenarios, such as granting permissions or handling complex page interactions, which broadens the skill's capability surface.
[Indirect Prompt Injection Surface]: By interacting with and reading content from arbitrary web pages, the skill processes untrusted data. This represents a security consideration where content from a visited website could potentially influence the agent's behavior.
Ingestion points: Data enters the agent's context through playwright-cli snapshot, eval results, and page content accessed via run-code.
Boundary markers: The skill does not currently specify explicit delimiters to distinguish between instructional content and data retrieved from the web.
Capability inventory: The skill has access to shell execution via Bash (playwright-cli, npx, npm) and the ability to read/write local files related to browser state.
Sanitization: There are no documented steps for sanitizing or filtering the content retrieved from external websites before it is processed by the agent.

playwright-cli