browser-testing-with-devtools
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: Static detections for instruction override phrases were identified as false positives. The skill includes phrases such as 'Ignore previous instructions' only as examples of untrusted content that the agent must specifically ignore when encountered in browser data.
- [EXTERNAL_DOWNLOADS]: The skill recommends installing the 'chrome-devtools-mcp' package via npx. This is a standard and well-known tool used for establishing a Model Context Protocol connection with browser developer tools.
- [DATA_EXFILTRATION]: The skill establishes strict security boundaries, instructing the agent to never copy-paste secrets or tokens found in browser content and forbidding external network requests from within the browser context.
- [COMMAND_EXECUTION]: JavaScript execution within the browser is explicitly restricted to read-only state inspection. The instructions prohibit credential access and require explicit user confirmation for any actions that modify page state.
Audit Metadata