skills/antfu/design/antfu-design/Gen Agent Trust Hub

antfu-design

Pass

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: SAFEPERSISTENCEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PERSISTENCE]: The skill instructs the agent to persist a design system rule by appending it to the project's AGENTS.md file if it is not already present. This ensures that future agent instances operating on the project adhere to the '@antfu/design' reuse guidelines.
  • [INDIRECT_PROMPT_INJECTION]: The skill describes an 'agent bridge' feature (@antfu/design/composables/agent) that allows agents to take UI snapshots and perform actions. This creates an indirect prompt injection surface where untrusted content displayed in the UI could potentially influence agent behavior when a snapshot is processed.
  • Ingestion points: UI components captured via agent.snapshot() or the ui_snapshot tool.
  • Boundary markers: None explicitly described in the documentation.
  • Capability inventory: The bridge allows agents to perform actions via agent.act() or ui_act, which can include triggering component-specific logic (e.g., form submissions, tab changes).
  • Sanitization: No specific sanitization or escaping of UI content before agent processing is mentioned.
  • [COMMAND_EXECUTION]: The documentation references standard CLI commands for environment setup and project maintenance.
  • Setup: pnpm add @antfu/design unocss and Storybook installation commands.
  • Documentation: pnpm docs:gen and pnpm run docs:overview are mentioned as methods to synchronize the component catalog and overview pages.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 20, 2026, 10:27 PM