The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses terminal() for directory creation and browser_exec() for browser automation, including opening local files and capturing screenshots for visual verification.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests user input for design preferences and references, which are then written into HTML files and opened in a browser environment via browser_exec. This creates a path for potentially malicious user input to execute within the browser context.\n
Ingestion points: User-provided 'Feel', 'References', and 'Core action' inputs collected during the intake phase.\n
Boundary markers: No delimiters or warnings are used when interpolating user content into HTML templates.\n
Capability inventory: terminal, write_file, and browser_exec (with local file access).\n
Sanitization: There is no mention of escaping or sanitizing the user-provided data before it is included in the mockup files.\n- [EXTERNAL_DOWNLOADS]: The generated mockups include styling from the Tailwind CSS library loaded from https://cdn.tailwindcss.com. This is a well-known service for prototyping that involves fetching external resources.\n- [REMOTE_CODE_EXECUTION]: The skill refers to the get-shit-done-cc package which is executed via npx. This indicates an external dependency on code from a public registry.

sketch