skills/hk-hub/agentskills/ui-design/Gen Agent Trust Hub

ui-design

Fail

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to modify the user's application source code by injecting a remote JavaScript file (https://ui.sh/ui-picker.js) into core layout files, such as index.html, RootLayout.tsx, app.vue, and app.blade.php. This facilitates the persistent execution of remote code within the application's context.
  • [COMMAND_EXECUTION]: The skill utilizes the npx @tailwindcss/cli canonicalize command to optimize CSS class strings. This involves running an external CLI tool that may download and execute code locally during the build or development process.
  • [EXTERNAL_DOWNLOADS]: The skill fetches assets and scripts from external domains, specifically ui.sh and assets.ui.sh. It also references third-party font services like Google Fonts and Fontshare for typography implementation.
  • [PROMPT_INJECTION]: The skill processes user-supplied images (screenshots, mockups) to generate code scaffolds in markup-from-image.md, creating a surface for indirect prompt injection.
  • Ingestion points: markup-from-image.md accepts screenshots and mockups for code generation.
  • Boundary markers: There are no explicit instructions for the agent to ignore or sanitize instructions that may be visually embedded in the processed images.
  • Capability inventory: The agent has the capability to write the resulting code directly to files in the repository and execute shell commands.
  • Sanitization: The skill lacks validation or filtering for text content extracted from processed images.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 19, 2026, 05:30 PM
Security Audit — agent-trust-hub — ui-design