@json-render/next

The skill allows defining server-side asynchronous functions (`loaders`) that execute arbitrary code before rendering. If an attacker can influence the `spec` (e.g., via AI prompt injection), they can inject malicious code into these loaders, leading to command execution or data exfiltration.

loaders: { loadPost: async ({ slug }) => { const post = await getPost(slug as string); return { post }; } }

The skill is designed for 'AI-generated multi-page applications' and provides a `schema` for 'AI catalog generation'. This implies an AI might produce the `NextAppSpec`, creating a supply chain vulnerability where a compromised AI could generate malicious `loaders` or configurations.

description: 'creating AI-generated multi-page applications.', 'schema -- Custom schema for Next.js apps (for AI catalog generation)'

The skill allows defining and manipulating global application state via the `state` field and built-in actions (`setState`, `pushState`, `removeState`). If an attacker can inject into the `spec` or control action parameters, they could poison the application's state, leading to data corruption or unintended behavior.

`state`: Global initial state shared across all routes, Built-in Actions: `setState`, `pushState`, `removeState`