The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill integrates with well-known frontend development tools, specifically referencing npx shadcn@latest init to fetch UI presets. This involves downloading configuration and code from a recognized third-party registry as part of standard developer workflows.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external configuration files including .tokens.json and components.json, which presents a potential surface for indirect injection if these files contain malicious instructions.
Ingestion points: SKILL.md describes reading components.json to determine active styles, and rules/tokens-w3c-format.md details the parsing of W3C-compliant token JSON files.
Boundary markers: The instructions do not define explicit boundary markers or instructions for the agent to ignore natural language found within the JSON values or descriptions.
Capability inventory: The skill utilizes Read, Glob, Grep, WebFetch, and WebSearch to interact with files and remote documentation.
Sanitization: The skill relies on standard JSON parsing and Style Dictionary transforms, without additional sanitization logic for data-driven prompt interpolation.

design-system-tokens