The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches UI components, themes, and templates from the 21st.dev registry. It also downloads SVG brand logos from the svgl.app library. These are standard operations for the tool's stated purpose.
[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of remote code into the local project using the 21st add command and npx shadcn@latest. This includes downloading component source files and installing associated npm dependencies.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the 21st CLI (@21st-dev/cli) for catalog searching, metadata retrieval, and project initialization.
[CREDENTIALS_UNSAFE]: The skill describes handling sensitive credentials, including API keys (21st_sk_...) and authentication tokens stored in ~/.config/21st/auth.json. These are handled through official vendor CLI mechanisms for authentication.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because it ingests and processes untrusted code and metadata from an external registry.
Ingestion points: External data enters the agent's context through 21st search results, 21st get (component code), and 21st logo (SVG URLs).
Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore instructions that might be embedded within the retrieved component code.
Capability inventory: The skill possesses the capability to write files to the local filesystem, install packages, and execute shell commands.
Sanitization: There is no explicit requirement for the agent to sanitize or validate the retrieved third-party code before integrating it into the project.

21st-cli-use