The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the vendor-provided 21st CLI tool to perform various operations including environment initialization (21st init), component searching (21st search), and UI variant generation (21st generate). These actions are consistent with the skill's stated purpose of UI exploration.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from both local and remote sources which could be used to influence the agent's output or execution flow.
Ingestion points: Data is ingested from the local .21st/design.json file and the output of the 21st search command.
Boundary markers: The instructions do not define explicit boundary markers or delimiters to isolate the ingested design data from the agent's system instructions.
Capability inventory: The agent has the ability to execute shell commands via the 21st CLI and record decisions back into the local .21st/design.json file.
Sanitization: There is no evidence of sanitization or schema validation performed on the design JSON or search results before they are incorporated into the agent's context.

21st-ui-explore