The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and execute the @google/design.md package. This is a trusted utility provided by Google used for linting and exporting design documentation.
[COMMAND_EXECUTION]: Shell commands are executed via npx to perform specific tasks: npx @google/design.md spec for specification checks, npx @google/design.md lint for structural validation, and npx @google/design.md export for generating output files.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted content from public URLs (DOM, styles, and CSS) and local source code repositories to reconstruct design systems.
Ingestion points: Publicly accessible websites in URL mode and local source files in Repository mode (SKILL.md).
Boundary markers: The instructions mandate a strict YAML/Markdown schema and require validation via an external linter, providing a structural boundary for processed data.
Capability inventory: The agent is authorized to write to the filesystem (DESIGN.md) and execute shell commands via npx (@google/design.md).
Sanitization: The skill requires mandatory structural linting and validation passes using the @google/design.md tool before any content is finalized or reported.

create-design-md