aesthetic

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • PROMPT_INJECTION (MEDIUM): Risk of Indirect Prompt Injection through the ingestion of external web content.
  • Ingestion Points: Workflow 1 uses the chrome-devtools skill to capture screenshots of untrusted external sites (Dribbble, Behance, Mobbin, etc.).
  • Boundary Markers: The instructions do not specify any delimiters or 'ignore embedded instructions' warnings for the analyzed content.
  • Capability Inventory: The skill integrates with ui-styling and web-frameworks, which involve writing code (Tailwind CSS, Next.js), and chrome-devtools, which can interact with browser sessions.
  • Sanitization: No sanitization or validation of the extracted design principles is mentioned before they are documented and potentially used to generate project files.
  • Risk: An attacker could place malicious instructions in images or metadata on 'inspiration' sites that the agent might interpret as legitimate design requirements or commands during the analysis phase.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 06:43 AM