widgets-ui
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- EXTERNAL_DOWNLOADS (HIGH): The skill instructs users to run
npx shadcn@latest add https://ui.inference.sh/r/widgets.json. This pattern downloads and executes code from a non-whitelisted external domain. - REMOTE_CODE_EXECUTION (HIGH): Usage of
npx skills addwith references such asinferencesh/skills@agent-uiallows for the installation and execution of arbitrary code from untrusted sources via a command-line interface. - COMMAND_EXECUTION (MEDIUM): Documentation provides shell commands that encourage users to execute installers and code-addition scripts directly in their environment from unverified 3rd party sources.
- INDIRECT_PROMPT_INJECTION (LOW): The skill renders UI widgets from 'agent responses'. This creates an attack surface where a compromised agent could generate UI that tricks the user (e.g., phishing forms or malicious action buttons). Evidence Chain: 1. Ingestion points:
WidgetRenderercomponentwidgetprop. 2. Boundary markers: Absent. 3. Capability: Form submission, button clicks, and UI rendering. 4. Sanitization: None visible in the provided snippets.
Recommendations
- AI detected serious security threats
Audit Metadata