The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation recommends using the shadcn CLI to fetch component configurations from https://ui.inference.sh/r/tools.json.
[COMMAND_EXECUTION]: Provides shell commands for installing the belt-sh/cli and various UI component packages via npx.
[PROMPT_INJECTION]: The UI components are designed to ingest and display untrusted data from tool execution cycles, creating an attack surface for indirect prompt injection.
Ingestion points: Data enters the UI context via the result prop in the ToolResult component and the args prop in ToolCall and ToolApproval components within SKILL.md.
Boundary markers: The provided documentation examples do not include explicit boundary markers or instructions for the agent to ignore potentially malicious embedded content in the results.
Capability inventory: The skill's primary capability is rendering provided data into the application's DOM for user display.
Sanitization: There is no evidence of sanitization or output escaping logic in the provided code snippets to prevent the rendering of malicious payloads within tool results.

tools-ui