skills/opencue/skills/gpt-taste/Gen Agent Trust Hub

gpt-taste

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [SAFE]: The skill focuses on defining aesthetic and layout constraints for AI-generated React code, ensuring visual quality and variety through design principles.
  • [EXTERNAL_DOWNLOADS]: References to picsum.photos for placeholder images and libraries such as GSAP and Phosphor Icons are safe and involve well-known, legitimate services.
  • [PROMPT_INJECTION]: The use of imperative language like 'CORE DIRECTIVE' and 'FORBIDDEN' is employed for style enforcement and defining the agent persona rather than attempting to subvert platform safety measures.
  • [REMOTE_CODE_EXECUTION]: Instructions to 'simulate a Python script execution' are intended for internal model reasoning to generate deterministic variety and do not represent a technical shell or system execution risk.
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface: 1. Ingestion points: User prompts (SKILL.md); 2. Boundary markers: Absent; 3. Capability inventory: Generates React UI code with external asset links and GSAP animations (SKILL.md); 4. Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 10:52 PM
Security Audit — agent-trust-hub — gpt-taste