The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Node.js script, scripts/storybook-to-catalog.mjs, to convert Storybook manifests into component catalogs. This is a standard developer workflow and involves local file operations only.
[DYNAMIC_EXECUTION]: The script scripts/storybook-to-catalog.mjs generates TypeScript source code (catalog.ts and components.tsx) based on an input JSON manifest. While it includes sanitization to prevent breaking out of string literals (escaping backslashes and single quotes in enums), it acts as a code generator which is a form of dynamic execution. The script includes a validation step to ensure generated Zod schemas do not contain z.any() or z.unknown(), maintaining the security posture of the framework.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the rendering of UI based on AI-generated JSON or YAML specs. This represents a potential surface for indirect prompt injection if external data is included in the specs.
Ingestion points: The Renderer component (documented in SKILL.md and rules/catalog-definition.md) accepts a spec object which defines the UI elements and props.
Boundary markers: The skill uses a 'Catalog' defined with Zod schemas as a hard boundary. SKILL.md states that if a component or prop is not in the catalog, the AI cannot generate it.
Capability inventory: The renderers (React, Vue, Svelte, etc.) can render UI components, handle events via the on field, and perform state mutations.
Sanitization: All props in the spec are validated at runtime against the Zod schemas defined in the catalog (rules/prop-constraints.md).
[EXTERNAL_DOWNLOADS]: The skill documentation and scripts reference external resources from Vercel Labs' GitHub organization. These are recognized as trusted sources and the references are documented neutrally.

json-render-catalog