design-loop

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads HTML components and screenshots from the Google Stitch SDK service using curl as part of the visual design generation process.- [COMMAND_EXECUTION]: Shell commands are used for project orchestration, including dependency management (npm), environment checks (ls), and running a local development server (npx serve) for verification.- [PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it reads instructions from a local state file (.design/next-prompt.md) to drive its autonomous build loop.
  • Ingestion points: The .design/next-prompt.md file is the primary source of instructions for each iteration.
  • Boundary markers: There are no delimiters or markers used to isolate instructions from the baton file content.
  • Capability inventory: The skill can execute Bash commands, perform network downloads via curl, and write/edit local files.
  • Sanitization: The skill does not implement validation or sanitization of the baton file content before it is processed by the AI models.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 08:36 AM