The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform project initialization and component installation using official CLI tools, specifically bunx shadcn-svelte@latest init and bunx shadcn-svelte@latest add. These commands are executed to automate the integration of UI libraries into the user's project.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and components from the shadcn-svelte registry and references official documentation at https://www.shadcn-svelte.com and https://svelte.dev. These are well-known, industry-standard resources for the Svelte development ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection as it processes user-provided design ideas and mockups while having the capability to write code and execute shell commands.
Ingestion points: User-supplied design descriptions, content density requirements, and brand direction parameters are processed in the workflow (SKILL.md).
Boundary markers: The skill utilizes a strict tool-based verification workflow, requiring the agent to use specific MCP tools (shadcn-svelte-search, shadcn-svelte-get) to ground its decisions in verified data before generating code.
Capability inventory: The agent is authorized to write SvelteKit files and execute package manager commands (references/mcp-workflow.md).
Sanitization: The instructions include explicit anti-hallucination rules that prohibit the invention of component names, props, or CLI commands, and mandate cross-referencing with official documentation.

shadcn-sveltekit-design