chrome-dev-mcp

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The script scripts/install-uxc.zsh downloads the uxc binary from a third-party GitHub repository (holon-run/uxc). Although the skill uses pinned SHA-256 checksums to verify the integrity of the downloaded archive and the extracted binary, the source is not a pre-verified trusted vendor.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes signals and data from the browser (e.g., page lists, Console output) which could contain adversarial content.
  • Ingestion points: scripts/uxc-readiness.zsh reads the standard output of the uxc CLI tool.
  • Boundary markers: The skill instructions in SKILL.md direct the agent to return sanitized sentinel values like CHROME_DEV_MCP_READY instead of raw tool output.
  • Capability inventory: The skill can execute shell scripts and interact with browser sessions via the DevTools protocol.
  • Sanitization: scripts/uxc-readiness.zsh utilizes jq to extract and validate specific JSON fields, filtering out unexpected data from the tool's response.
  • [DYNAMIC_EXECUTION]: scripts/setup-uxc-link.zsh dynamically generates a wrapper shell script (chrome-dev-mcp-cli) that incorporates local configuration such as exclusive keys and directory paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 02:26 AM
Security Audit — agent-trust-hub — chrome-dev-mcp