The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the shadcn CLI via package runners (npx, pnpm, bun) to manage the UI development lifecycle. It also utilizes dynamic context injection in SKILL.md via the command ``!`npx shadcn@latest info --json```, which executes at load time to populate project metadata such as Tailwind configuration and import aliases.
[EXTERNAL_DOWNLOADS]: The instructions require the agent to fetch component documentation, code examples, and registry manifests from well-known services, specifically ui.shadcn.com and GitHub repositories. These operations are consistent with the documented features of the shadcn/ui framework.
[PROMPT_INJECTION]: The skill ingests untrusted data from external registry items and documentation URLs, which constitutes a surface for indirect prompt injection.
Ingestion points: Registry item definitions and documentation pages fetched via the CLI (SKILL.md, cli.md).
Boundary markers: The skill does not explicitly define delimiters to isolate external content from its core instructions.
Capability inventory: The skill has access to the Bash tool for running CLI commands and file system access for writing UI components (SKILL.md, cli.md).
Sanitization: The skill relies on the standard parsing of registry JSON files but does not mention specific sanitization for the narrative documentation content.

shadcn