ui-designer

Warn

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Several operations involve executing shell commands with arguments provided by the user, such as the port number in bash scripts/setup.sh <port>, the project name in zip -r /tmp/<project>.zip <project>/, and image paths in bash scripts/convert-image.sh <input> <output.webp>. This creates a surface for command injection if input is not properly sanitized.
  • [EXTERNAL_DOWNLOADS]: The UI template used by the skill fetches resources from external CDNs, including Tailwind CSS (cdn.tailwindcss.com) and React libraries (unpkg.com). Additionally, the installation command retrieves the skill's source from a remote GitHub repository.
  • [PROMPT_INJECTION]: The skill translates natural language user prompts into React code that is subsequently rendered for visual review. This creates an indirect prompt injection surface where a malicious prompt could result in the execution of arbitrary JavaScript within the review environment.
  • Ingestion points: Natural language UI prompts in SKILL.md workflow.
  • Boundary markers: Absent.
  • Capability inventory: Bash script execution, zip command execution, and headless browser screenshotting.
  • Sanitization: Absent; the skill relies on direct translation of user intent into code without specified escaping or filtering.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 26, 2026, 01:40 AM
Security Audit — agent-trust-hub — ui-designer