ape-design
Fail
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill mandates that the agent configure a custom registry (https://ui.arpitbhayani.me/r/{name}.json) and execute the shadcn CLI with the '-y' flag to install components. This process allows the agent to download and integrate external code into the project environment without manual user review or approval.
- [COMMAND_EXECUTION]: The instructions explicitly command the agent to run shell commands, including 'npm install array-ui' and 'npx shadcn@latest add @aui/...', to manage project dependencies.
- [PROMPT_INJECTION]: The skill contains instructions that override standard safety protocols, specifically stating that the agent 'MUST NOT ask the user which components to install' and should act 'autonomously.' This suppresses the user's ability to review and approve potentially dangerous shell commands and code downloads.
- [EXTERNAL_DOWNLOADS]: The skill fetches fonts from 'fonts.googleapis.com' and component configurations from 'ui.arpitbhayani.me', which is a domain that closely mimics the real developer's identity but is associated with a suspicious author account (arpitbbhayani).
Recommendations
- AI detected serious security threats
Audit Metadata