chrome-cdp

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/cdp.mjs uses child_process.spawn to launch itself as a background daemon process. This is used to maintain persistent browser sessions and follows a standard pattern for CLI tools of this type.
  • [DATA_EXPOSURE]: The script reads browser metadata files (such as DevToolsActivePort) and environment variables to discover the local Chrome debugging port. Access is limited to standard browser configuration paths and local system metadata.
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it retrieves and processes content from arbitrary web pages.
  • Ingestion points: Raw HTML and accessibility tree data are retrieved from external websites via scripts/cdp.mjs using the html and snap commands.
  • Boundary markers: The instructions do not define specific delimiters for separating web content from system instructions, although the SKILL.md provides behavioral guidance.
  • Capability inventory: The skill has significant interaction capabilities, including navigation (nav), element interaction (click, type), and arbitrary browser-side script execution (eval).
  • Sanitization: No automated sanitization of retrieved page content is performed by the tool scripts.
  • [EXTERNAL_DOWNLOADS]: The skill connects to a local browser instance via HTTP and WebSockets on 127.0.0.1. No remote external downloads or unauthorized network connections were identified in the codebase.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 03:27 PM
Security Audit — agent-trust-hub — chrome-cdp