visual-verify

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill initiates local development environments using project-defined commands such as npm run dev, docker compose up, or just up. It also performs local network connectivity checks using curl against common development ports.
  • [EXTERNAL_DOWNLOADS]: If a browser is not present, the skill installs Chromium using the npx playwright install command, which fetches the binary from the official Playwright distribution.
  • [PROMPT_INJECTION]: The skill is designed to process untrusted external data (rendered web pages and PDFs). To mitigate potential indirect prompt injection, the instructions strictly limit the agent to a 'Capture, Look, Articulate, Report' workflow and explicitly prohibit implementing any code fixes or changes based on the visual review.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 07:18 AM
Security Audit — agent-trust-hub — visual-verify