The Agent Skills Directory

Terminal UI Rendering from JSON Specifications: The skill allows agents to define and render complex terminal interfaces using a JSON element map. This is a common pattern for creating interactive CLI tools and does not pose a direct security risk when used as intended.
Indirect Prompt Injection Surface: As a UI renderer, the skill ingests and processes JSON specifications to generate output. There is a potential consideration where an untrusted specification could be used to present misleading information or spoof UI elements to a user. Developers should ensure that UI specifications from untrusted sources are appropriately validated.
Evidence Chain for Data Processing:\n
Ingestion points: The Renderer component and useUIStream hook process JSON data to build the UI tree.\n
Boundary markers: The current documentation does not specify explicit boundary markers for the renderer.\n
Capability inventory: The skill facilitates UI interactions (inputs, selects, links) and state management within the terminal environment.\n
Sanitization: Standard JSON parsing is used; specific UI-level sanitization for preventing injection via templates is a recommended consideration for users of the library.

ink