The Agent Skills Directory

[PROMPT_INJECTION]: The skill documents components for rendering message content, such as MessagePrimitive.Parts, which represents a potential surface for indirect prompt injection. If an attacker can influence the content of messages rendered by these components, they might attempt to inject instructions aimed at the agent processing the conversation.
Ingestion points: MessagePrimitive.Parts (documented in SKILL.md and references/message.md) handles the display of text and tool results.
Boundary markers: The documentation does not specify explicit boundary markers or 'ignore' instructions for the rendered parts.
Capability inventory: The skill is documentation-focused and does not include scripts for subprocess execution, local file system modification, or persistent network connections.
Sanitization: The provided examples do not demonstrate sanitization, which is expected to be managed by the implementation's runtime environment.
[EXTERNAL_DOWNLOADS]: The documentation references official resources from assistant-ui.com and its associated registry at r.assistant-ui.com. These references are used for fetching library-specific configuration and are standard for this technology.
[SAFE]: The skill adheres to standard React development patterns, utilizing well-known libraries and local API endpoints (e.g., /api/chat) for its examples.

primitives