The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from user-provided directories and files, which could contain instructions intended to influence the agent's behavior.
Ingestion points: Reference images are processed in Step 2, and a project idea file is ingested in Step 3 of the workflow in SKILL.md.
Boundary markers: The skill utilizes XML-like tags (e.g., <context>, <goal>) and Markdown headers within its template files (assets/design-system.md, assets/app-overview-generator.md) to delimit data, which helps mitigate but does not eliminate injection risks.
Capability inventory: The skill can read local files, execute shell commands (npm, find), and write new React components to the file system.
Sanitization: No explicit sanitization or validation logic is defined for the input text or image data before it is interpolated into subagent prompts.
[COMMAND_EXECUTION]: The skill executes shell commands to check the environment and install necessary development tools.
Evidence: Uses find . -name "package.json" -exec grep -l "react" {} \; to verify the existence of a React project in SKILL.md.
Evidence: Recommends executing npx create-react-app, npm install, and npx tailwindcss init for project initialization.
Context: These commands are standard for modern web development workflows and are used according to their intended purpose.

ui-designer