The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of UI components and CLI tools from remote repositories using npx shadcn and npx skills. Specifically, it fetches configuration from https://ui.inference.sh/r/chat.json and installs skills from the belt-sh and inference-sh namespaces. These resources are consistent with the skill's stated purpose of providing UI building blocks.
[INDIRECT_PROMPT_INJECTION]: The skill defines components for rendering chat messages, creating an attack surface where maliciously crafted data from an LLM response could attempt to influence the agent's behavior if the consuming application does not properly sanitize inputs.
Ingestion points: Untrusted message content passed to the content property of ChatMessage components in SKILL.md.
Boundary markers: Absent in the UI-focused code snippets.
Capability inventory: No tool execution or sensitive file access capabilities are defined in this skill.
Sanitization: None present in the presentation-layer code provided.

chat-ui