chrome-devtools-mcp
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
npx -y chrome-devtools-mcp@latestto download and run the MCP server from the public npm registry during the setup process in Step 2. - [COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands for environment setup and troubleshooting:
claude mcp add chrome-devtools ...to configure the MCP server.npm cache clean --forcefor troubleshooting package issues./Applications/AutoClickCDPPopup.app/Contents/MacOS/auto-click-cdp-popupto check status and request bridge operations when the standard CDP connection fails.- [DYNAMIC_EXECUTION]: The tool
mcp__chrome-devtools__evaluate_scriptallows for the execution of arbitrary JavaScript functions within the context of a connected Chrome browser tab (e.g.,function="() => document.title"). - [INDIRECT_PROMPT_INJECTION]: The skill has the surface area for indirect prompt injection because it ingests untrusted data from the web.
- Ingestion points: Browser content is retrieved via
mcp__chrome-devtools__take_snapshotandmcp__chrome-devtools__list_pages(SKILL.md). - Boundary markers: The instructions do not define explicit delimiters or warnings to ignore instructions embedded within the retrieved DOM content.
- Capability inventory: The skill can execute shell commands (
claude mcp add) and browser-based scripts (evaluate_script). - Sanitization: There is no evidence of sanitization or filtering applied to the DOM data before it is processed by the agent.
Audit Metadata