The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the download of a component manifest from https://ui.inference.sh/r/agent.json via the shadcn CLI tool.
[EXTERNAL_DOWNLOADS]: The documentation recommends installing the @inferencesh/sdk package to handle API proxying for the agent runtime.
[COMMAND_EXECUTION]: Provides shell commands for environment setup, including npx shadcn to add the component and npx skills to add related blocks like chat-ui and widgets-ui.
[PROMPT_INJECTION]: The skill builds a UI that processes untrusted user input and can execute client-side tools (scan_ui, fill_field), representing an indirect prompt injection surface.
Ingestion points: User messages and system prompts processed by the Agent component.
Boundary markers: None explicitly detailed in the setup documentation; users should implement their own delimiters.
Capability inventory: Execution of client-side tools and network proxying for model interactions.
Sanitization: No specific sanitization or validation logic is provided in the skill's setup instructions.

agent-ui