The Agent Skills Directory

[Command Execution] (SAFE): The skill executes a find command in Step 5 to detect existing React projects by searching for 'react' within package.json files. This is a standard read-only environment check.
[External Downloads] (LOW): Step 5 recommends running npm install and npx to set up Tailwind CSS and React dependencies. These installations target the official npm registry, which is a trusted source, and are necessary for the skill's primary purpose.
[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted external data including user-provided UI screenshots and project idea documents.
Ingestion points: Reference images directory and project idea markdown files.
Boundary markers: Prompt templates in assets/ utilize XML-style tags (e.g., <context>, <goal>) to delimit user data from system instructions.
Capability inventory: Subprocess execution for shell commands (find) and package management (npm).
Sanitization: There is no explicit sanitization or filtering of the text extracted from images or provided in project documents before they are interpolated into subagent prompts.

ui-designer