The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill accepts user input for design vibes, references, and content, which it then interpolates into HTML mockup files. These files are subsequently loaded and analyzed by the agent using vision tools, creating a potential surface for injection.
Ingestion points: User-provided descriptions for design "feel", "references", and "realistic fake content" are incorporated into the HTML templates.
Boundary markers: No specific delimiters or warnings are used to instruct the agent to ignore instructions that might be embedded in the user-provided content.
Capability inventory: The skill uses terminal, write_file, browser_navigate, and browser_vision to create and inspect the mockups.
Sanitization: The instructions do not include steps for sanitizing or escaping user-provided text before it is written to the HTML files and executed in the browser environment.
[EXTERNAL_DOWNLOADS]: The skill documentation references an external npm package (get-shit-done-cc) and suggests its installation via npx, which involves downloading and executing code from the npm registry.
[COMMAND_EXECUTION]: The skill utilizes the terminal and write_file tools to manage a local sketches/ directory, which is necessary for its core function of generating design variants.
[SAFE]: The skill loads the Tailwind CSS utility library from a well-known service (https://cdn.tailwindcss.com) to provide styling for the generated mockups.

sketch