The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and utilizes tools, specifications, and configurations from trusted organizations and well-known technology services. Each instance is handled as a standard dependency or reference.
Fetches and uses the DESIGN.md specification and CLI from a repository associated with Google (github.com/google-labs-code/design.md).
References design system integration techniques provided by Vercel Labs (github.com/vercel-labs/design-systems-to-agent-skills).
Recommends industry-standard development and testing tools including axe-core, Playwright, ESLint, and Stylelint.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external brand tokens and design specifications, which constitutes a vulnerability surface where malicious instructions could be embedded in the data files.
Ingestion points: Processes DESIGN.md (YAML/Markdown) and tokens.json files to guide the agent's UI generation tasks as described in references/design-md-and-enforcement.md.
Boundary markers: The skill encourages structured YAML for tokens but does not explicitly implement delimiters to isolate instructions from data within these files.
Capability inventory: The skill possesses capabilities for UI code generation, writing files to the repository, and executing CLI tools for linting and visual testing.
Sanitization: The skill advocates for technical enforcement mechanisms such as TypeScript discriminated unions and strict linting rules like no-arbitrary-colors to validate inputs and prevent drift.

visual-design-system