design-style

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is instructed to generate a comparison HTML file and save it to a specific local path: ~/Desktop/Claude skills/design-style-preview-<project-slug>.html. It then triggers the open command to launch this file in the default web browser (SKILL.md Phase 2.5).
  • [COMMAND_EXECUTION]: To integrate design styles into an existing project, the skill reads local source files including HTML, CSS, and framework-specific configurations such as Tailwind CSS files (SKILL.md Phase 4.2).
  • [PROMPT_INJECTION]: The skill ingests user-supplied text regarding project purpose and brand vibe. This data enters the agent's context and influences subsequent code generation and file content without explicit boundary markers or sanitization, representing an indirect prompt injection surface (SKILL.md Phase 1, Phase 2).
  • [SAFE]: The skill references an external interactive reference at design-lab-yanliu.vercel.app. This is a vendor-owned resource hosted on a well-known service provider, used solely for design reference.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 07:39 AM
Security Audit — agent-trust-hub — design-style