The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches component configuration and related building blocks from the official ui.inference.sh and inference-sh repositories using npx shadcn and npx skills commands.
[EXTERNAL_DOWNLOADS]: Installs the vendor's official SDK (@inferencesh/sdk) from the npm registry.
[CREDENTIALS_UNSAFE]: Instructions correctly recommend storing sensitive API keys (INFERENCE_API_KEY) in a .env.local file, which is the standard safe practice for environment variable management.
[PROMPT_INJECTION]: The skill implements a chat interface with 'human-in-the-loop' features and client-side tools like scan_ui and fill_field. This creates an indirect prompt injection surface where untrusted data from scanned web pages or UI elements could attempt to manipulate the agent's behavior.
Ingestion points: User input through the Agent component and automated UI data through the scan_ui tool (SKILL.md).
Boundary markers: None explicitly documented in the provided implementation examples.
Capability inventory: Includes client-side tool execution, form filling, and human-in-the-loop approval flows (SKILL.md).
Sanitization: No specific sanitization or escaping of scanned UI content is detailed in the component configuration.

agent-ui