frontend-audit

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Downloads OmniParser-v2.0 model weights from Microsoft's official repository on Hugging Face during the bootstrap phase.
  • [EXTERNAL_DOWNLOADS]: Installs standard image processing and browser automation dependencies from NPM and PyPI registries.
  • [COMMAND_EXECUTION]: Executes shell commands to manage a local Python virtual environment, install system packages (via apt or brew), and run image analysis scripts.
  • [COMMAND_EXECUTION]: Employs Playwright to capture screenshots and extract computed CSS styles from a user-specified development server.
  • [PROMPT_INJECTION]: Indirectly processes content from a local development URL. Ingestion: Playwright DOM extraction in check.mjs. Boundary markers: None. Capabilities: Bash script execution and file system writes. Sanitization: Employs CSS selector escaping for locator stability.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 12:58 PM
Security Audit — agent-trust-hub — frontend-audit