The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation includes commands to download a component registry from https://ui.inference.sh/r/agent.json via the shadcn CLI during setup.
[COMMAND_EXECUTION]: The skill requires the execution of several shell commands for installation and dependency management, including npm install @inferencesh/sdk, npx shadcn add, and multiple npx skills add calls for related components.
[INDIRECT_PROMPT_INJECTION]: The skill implements a chat interface and agentic UI that ingests untrusted data from external LLM providers and client-side tools, creating a potential vector for indirect prompt injection.
Ingestion points: Agent responses, token streams, and widget data processed by the <Agent /> component (found in SKILL.md).
Boundary markers: No explicit boundary markers or "ignore instructions" directives are specified in the provided usage examples.
Capability inventory: The component supports file/image uploads and execution of browser-based tools (e.g., scan_ui, fill_field) that interact with the application state.
Sanitization: The provided instructions do not include specific steps for sanitizing or validating LLM-generated output before UI rendering or tool execution.
[NO_CODE]: This skill consists entirely of documentation and usage instructions in markdown format; it does not distribute any executable scripts or binaries within its own package files.

agent-ui