The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file uses the dynamic context injection syntax (!) to execute npx shadcn@latest info --json when the skill is loaded. This is a functional use of the platform's features to provide the agent with immediate access to project details like import aliases, framework version, and installed components.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch external content, such as documentation and examples, from URLs generated by the npx shadcn@latest docs command. Because this external data is ingested into the agent's context at runtime, it represents an indirect prompt injection surface. The skill includes specific instructions for the agent to review and verify all added files, which acts as a manual verification step.
[COMMAND_EXECUTION]: The primary function of the skill involves executing the shadcn CLI to manage project files. This includes adding new components and updating existing ones. To minimize the risk of unintended modifications, the instructions mandate the use of preview flags like --dry-run, --diff, and --view before applying changes.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of registry metadata and component source code from external sources, including the official shadcn registry and public GitHub repositories (owner/repo/item). These operations are handled via the CLI's registry resolution system and are standard for the framework's operation.

shadcn