Playwright MCP Tester
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
@playwright/mcppackage using package managers likenpmorbun. It also suggests running the server vianpx @playwright/mcp@latest. - [COMMAND_EXECUTION]: The testing workflow involves the agent executing shell commands to start local development servers (e.g.,
bun dev). - [DYNAMIC_EXECUTION]: The tool
playwright_evaluateprovides the capability to execute arbitrary JavaScript code within the browser context. This is a powerful feature that allows deep interaction with web pages but could be misused to access browser-based secrets like cookies or local storage if the agent is directed to sensitive sites. - [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes content from external web applications. Malicious text or instructions embedded in a website's UI could potentially override the agent's intended testing logic.
- Ingestion points: Web page content and screenshots retrieved during the test loop in
SKILL.md. - Boundary markers: None present; the agent analyzes raw screenshots and DOM state without specific delimiters.
- Capability inventory: Shell command execution, file system modification, and browser-based JavaScript execution.
- Sanitization: No mechanisms for sanitizing or filtering instructions found within the tested web applications are described.
Audit Metadata