The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a bash script scripts/check-versions.sh that executes npm commands to query package metadata.
Evidence: The script runs npm list and npm view to verify the local versions of ai, @ai-sdk/openai, and other dependencies against the npm registry.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting and rendering data provided by external AI models.
Ingestion points: templates/nextjs-chat-app-router.tsx, templates/custom-message-renderer.tsx, and templates/use-chat-tools.tsx (all process the messages array containing model output).
Boundary markers: Absent; templates do not implement specific delimiters or instructions for the agent to ignore potentially malicious instructions embedded in model responses.
Capability inventory: The skill has local storage access via localStorage in templates/message-persistence.tsx and can make outbound API calls via the useChat hook.
Sanitization: Employs react-markdown in the custom renderer template, though it provides a commented-out example that utilizes dangerouslySetInnerHTML.

ai-sdk-ui