The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the dynamic context injection pattern !npx shadcn@latest info --json in the SKILL.md file to automatically retrieve project settings (framework, Tailwind version, etc.) when the skill is loaded. This execution is benign and used to tailor the agent's behavior to the specific project.
[EXTERNAL_DOWNLOADS]: The skill frequently uses the shadcn CLI to download UI components and documentation from the official registry (ui.shadcn.com) and supports adding components from custom or community-defined URLs and registries.
[PROMPT_INJECTION]: The skill instructs the agent to fetch documentation and examples from URLs provided by the CLI and to process code from external registries, introducing a surface for indirect prompt injection.
Ingestion points: Documentation URLs from npx shadcn@latest docs and source code from third-party community registries.
Boundary markers: The skill does not define specific delimiters or instructions to isolate or ignore embedded prompts in the fetched external content.
Capability inventory: The agent has the ability to write files to the project, install npm dependencies, and execute shell commands.
Sanitization: The instructions include a workflow step to "review added components" and "verify they are correct" as a manual safety check.

shadcn