chrome-devtools

Warn

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/install-deps.sh script utilizes sudo to install necessary system libraries for Chromium on Linux distributions (Ubuntu, Debian, Fedora, Arch). This is a privileged operation required for the tool's environment setup.
  • [REMOTE_CODE_EXECUTION]: The scripts/evaluate.js script allows for the execution of arbitrary JavaScript code within the browser context using the eval() function, which can be leveraged to interact with or manipulate web content dynamically.
  • [COMMAND_EXECUTION]: The SKILL.md instructions guide the agent to create and run custom Node.js automation scripts in a temporary project directory (.claude/chrome-devtools/tmp/), which involves local code generation and execution.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests untrusted data from the DOM, ARIA trees, and console logs of external websites which then influences the agent's context and subsequent actions.
  • Ingestion points: aria-snapshot.js, snapshot.js, console.js, and network.js read and format content from external URLs.
  • Boundary markers: Not explicitly enforced in prompt instructions to separate web data from agent commands.
  • Capability inventory: Includes clicking elements, filling forms, and executing arbitrary JavaScript via evaluate.js.
  • Sanitization: The skill includes a validateXPath function in scripts/lib/selector.js that attempts to block common injection patterns like javascript:, <script, and event handlers in selectors.
  • [EXTERNAL_DOWNLOADS]: The installation scripts (install.sh, install-deps.sh) download software packages and dependencies from official and well-known repositories such as the NPM registry and system package managers.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 3, 2026, 01:36 AM
Security Audit — agent-trust-hub — chrome-devtools