The Agent Skills Directory

[No Code] (SAFE): The skill consists entirely of Markdown documentation. No executable scripts, system commands, or configuration files were found across the provided files.
[Indirect Prompt Injection] (SAFE): The library described in the documentation is designed to process untrusted data (AI responses, user messages, and tool outputs), which is an inherent attack surface for chat interfaces.
Ingestion points: File references/architecture.md defines the ThreadMessage data structure which ingests text, images, and tool results.
Boundary markers: Not explicitly defined in this architectural overview.
Capability inventory: File references/packages.md lists support for markdown rendering and tool call handling.
Sanitization: The documentation mentions the safe-content-frame package specifically for sandboxed iframe content, indicating a focus on security for rendered content.

assistant-ui