The Agent Skills Directory

[SAFE]: The skill utilizes the @google/design.md tool for linting and references the corresponding specification. Both the tool and specification are provided by Google, a well-known and trusted organization.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via an interview process to generate design artifacts. This is an expected behavior for its primary purpose.
Ingestion points: User responses to the 8–12 questions in the adaptive interview flow defined in SKILL.md.
Boundary markers: No specific delimiters or "ignore instructions" markers are applied to user input before interpolation into the artifact generation phase.
Capability inventory: The skill produces local files (DESIGN.md, PRODUCT.md, and a preview HTML file).
Sanitization: The skill relies on agent-side logic to parse and translate conversational input into structured design tokens without explicit sanitization steps.

design-init