Design to Code

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (MEDIUM): The skill is susceptible to Indirect Prompt Injection from untrusted design files.
  • Ingestion points: Design data is ingested via mcp__pencil__batch_get and mcp__pencil__get_variables tool calls that read external .pen files.
  • Boundary markers: Absent. The instructions do not define delimiters or provide guidance to the agent to ignore or separate natural language instructions embedded within the design properties (e.g., text content).
  • Capability inventory: The skill is designed to generate executable React/TSX code. If the AI agent has permissions to modify or create files in the workspace, a maliciously crafted design file could trick the agent into writing backdoors or exfiltration scripts into the codebase.
  • Sanitization: Absent. The skill provides mapping tables for styling but does not include logic for sanitizing or validating text content from the design before it is interpolated into component templates.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 01:04 AM