browser-testing-with-devtools
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's setup instructions direct the user to configure the agent with a command that uses
npxto fetch thechrome-devtools-mcppackage from the NPM registry at runtime. - [COMMAND_EXECUTION]: The skill provides a shell command (
npx -y chrome-devtools-mcp@latest --isolated) intended for the agent to execute as a background service (MCP server) to interact with the Chrome browser. - [REMOTE_CODE_EXECUTION]: The recommended configuration uses
npxwith the@latestversion tag, which results in the download and execution of the most recent version of the tool from a remote repository whenever the server is started. - [INDIRECT_PROMPT_INJECTION]: The skill explicitly identifies and mitigates the risk of indirect prompt injection from malicious web pages.
- Ingestion points: Browser DOM content, console logs, network responses, and JavaScript execution results (SKILL.md).
- Boundary markers: The instructions mandate that agents must clearly label browser findings and never merge untrusted browser content into the trusted instruction context.
- Capability inventory: The skill utilizes tools for JavaScript execution, DOM inspection, and network monitoring.
- Sanitization: It establishes strict rules requiring user confirmation before navigating to URLs found in page content or performing page mutations.
Audit Metadata