chrome-devtools-mcp
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides the ability to process and interact with untrusted web content from external websites via tools like
take_snapshotandexecute_script. This is a standard vulnerability surface for browser automation tools where a website could theoretically include instructions meant for the AI agent. - Ingestion points: The skill navigates to external URLs and retrieves DOM structures through
take_snapshotandexecute_script. - Boundary markers: The instructions do not define specific boundary markers to isolate website content from instructions.
- Capability inventory: The skill has broad browser interaction capabilities including clicking, typing, script execution, and extension installation.
- Sanitization: No explicit sanitization of web content is described in the prompt logic.
- [COMMAND_EXECUTION]: The skill utilizes JavaScript execution within the browser context (
execute_script) to perform various tasks such as error logging, layout checking, and accessibility auditing. This is standard functionality for browser-based automation skills. - [EXTERNAL_DOWNLOADS]: The documentation references official documentation and troubleshooting guides from Chrome DevTools and the
chrome-devtools-mcpGitHub repository.
Audit Metadata