The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the browser automation server from Anthropic's official NPM registry using npx. This is a trusted source and follows standard deployment practices.
[COMMAND_EXECUTION]: Executes the npx command to launch the Playwright MCP server. This is the intended operation for providing browser automation features to the agent.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it is designed to ingest and process data from external websites.
Ingestion points: Tools such as navigate_page, get_page_content, and get_text defined in SKILL.md allow untrusted external data to enter the agent's context.
Boundary markers: The skill documentation does not mention the use of delimiters or specific instructions to help the model distinguish between system instructions and content retrieved from the web.
Capability inventory: The skill includes powerful interaction tools like click, fill, and evaluate (JavaScript execution) which could be misused if the agent follows instructions hidden within a malicious webpage.
Sanitization: There is no evidence of filtering or sanitizing the web content before it is processed by the agent.

playwright