The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates project initialization and maintenance by generating and executing standard development commands such as pnpm install, turbo run build, and style-dictionary build. These operations are well-scoped to the stated purpose of setting up a design system monorepo.
[CREDENTIALS_UNSAFE]: Security best practices are followed for credential management. The skill explicitly guides the use of .env files for local secrets and utilizes GitHub Actions secret placeholders (e.g., ${{ secrets.NPM_TOKEN }}) in generated CI/CD templates, ensuring sensitive tokens are not hardcoded.
[EXTERNAL_DOWNLOADS]: All external tools and dependencies referenced—including Radix UI, tsdown, Turborepo, and Storybook—are well-known, widely-adopted packages within the frontend ecosystem sourced from standard registries (NPM).
[PROMPT_INJECTION]: An indirect prompt injection surface exists in 'Enforced Mode' where the agent reads instructions from a local .design-system/system.md file. While this is the core intended functionality for state persistence, it represents a surface where untrusted modifications to that file could influence agent behavior.
Ingestion points: Reads project-specific rules from .design-system/system.md and processes user-provided UI inventory data in Phase 1.
Boundary markers: The skill expects a specific Markdown schema but does not include explicit warnings to ignore embedded instructions within the configuration file.
Capability inventory: The agent possesses file-writing capabilities, directory management, and the ability to generate React component code and CSS.
Sanitization: There is no explicit validation or sanitization of the configuration values before they are adopted as instructions.

design-system