The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a design system checker utility script that interacts with the system environment to analyze the codebase.
Evidence: reference/scripts/design-system/check-system.mjs uses node:child_process modules, specifically execSync and execFileSync to call the local git binary for repository introspection, and spawnSync to execute itself using the Node.js path for self-testing logic.
[EXTERNAL_DOWNLOADS]: The project references several external libraries for UI primitives, internationalization, and animation effects.
Evidence: reference/REFERENCE.md specifies dependencies including standard libraries like @base-ui/react, @radix-ui, and @tanstack/react-router, as well as specialized UI packages like blendy, torph, morphicons, and streamdown.
[INDIRECT_PROMPT_INJECTION]: The skill implements various user-facing forms and search fields which ingest data that could potentially influence an AI agent's logic if processed without adequate delimiters.
Ingestion points: SearchInput in reference/components/ui/forms/filters/search-input.tsx and AuthTextField in reference/components/auth/auth-form.tsx accept arbitrary string inputs from users.
Boundary markers: The components use standard React DOM rendering which provides inherent protection against HTML injection (XSS), but no explicit instructions or delimiters for AI safety (e.g., 'ignore embedded instructions') are implemented in the input components.
Capability inventory: The reference code includes file system read/write access and command execution capabilities via the check-system.mjs CI script.
Sanitization: Content is escaped via React's default rendering engine, effectively neutralizing standard web-based injection vectors like XSS.

app-ui