The Agent Skills Directory

Indirect Prompt Injection Surface: The skill interacts with and processes content from external websites via snapshots and DOM evaluation. This could potentially allow untrusted web content to influence the agent's behavior.
Ingestion Points: Data enters the agent context through snapshots and element evaluations from visited web pages.
Boundary Markers: No specific delimiters or instructions are provided to the agent to treat page content as untrusted.
Capability Inventory: The agent possesses capabilities to manage browser states, execute scripts, and control network routing.
Sanitization: Web content is ingested as-is without specific sanitization for prompt injection patterns.
Dynamic Browser Code Execution: Commands like run-code and eval allow for the execution of arbitrary JavaScript within the browser session. This is an essential feature for complex automation tasks but involves the dynamic execution of code.
Evidence: playwright-cli run-code is documented for scenarios like geolocation overrides and custom browser logic.
Sensitive Data Management: The skill provides commands to manage and export browser authentication states, such as cookies and local storage, to local files.
Evidence: playwright-cli state-save and playwright-cli cookie-get allow for the retrieval and storage of browser state data.
Environment Setup via External Tools: The documentation includes steps for installing and running the Playwright CLI tool from a public registry. This is a standard setup procedure for using the automation framework.
Evidence: The skill recommends using npm install -g @playwright/cli@latest and npx for installation and execution.

playwright-cli