The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill configuration uses npx to download and execute @playwright/mcp@latest. This package is an official tool provided by Microsoft.
[DYNAMIC_EXECUTION]: Includes the browser_evaluate tool, which allows the agent to execute custom JavaScript within the browser's execution context. This is a standard and intended feature for browser automation and testing.
[COMMAND_EXECUTION]: The setup process requires the execution of the npx command to initialize the server, which is a standard procedure for Node.js-based MCP servers.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to navigate and process content from external websites, which introduces a surface for indirect prompt injection if a visited site contains malicious instructions meant for the AI agent.
Ingestion points: Live web content and accessibility trees retrieved via browser_navigate.
Boundary markers: Not explicitly defined in the skill instructions; relies on the agent's internal safety filters.
Capability inventory: The agent can interact with the DOM via browser_click, browser_type, and browser_evaluate.
Sanitization: The skill does not provide explicit sanitization for web content, relying on the underlying Playwright framework and LLM guardrails.

playwright-mcp