makeui

Warn

Audited by Gen Agent Trust Hub on Oct 20, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The script scripts/verify-mobile-package-surface.mjs performs dynamic loading of modules using the import() function with paths computed at runtime. It specifically imports components from the @qfei-design/make-app-mobile package and executes code through renderToStaticMarkup to verify the HTML output of those components.\n- [INDIRECT_PROMPT_INJECTION]: This skill possesses a significant attack surface for indirect prompt injection.\n
  • Ingestion points: The skill consumes untrusted "host-provided object/field metadata" to generate UI structures and field controls (referenced in SKILL.md).\n
  • Boundary markers: The instructions mention a "normalized metadata boundary" and the use of "permission-derived field-set handoff" to limit visible fields, which acts as a partial boundary.\n
  • Capability inventory: The skill can generate source code files for React pages, execute shell commands for package installation, and run scripts that perform dynamic imports.\n
  • Sanitization: SKILL.md describes a normalization process for schema metadata before it is passed to components.\n- [COMMAND_EXECUTION]: The skill instructions and scripts involve running shell commands, specifically package manager operations like corepack pnpm add, to manage and verify the environment and its dependencies.\n- [EXTERNAL_DOWNLOADS]: The skill requires downloading several external packages from the NPM registry, including vendor-specific resources like @qfei-design/make-app-mobile and well-known libraries like lucide-react.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 20, 2026, 06:50 AM