The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/shadcn_add.py script executes the shadcn CLI using subprocess.run. This is a core functional requirement of the skill for programmatic component installation. The implementation is secure, using a list of arguments rather than a shell string, which prevents command injection vulnerabilities from user-supplied component names.\n- [EXTERNAL_DOWNLOADS]: The skill uses npx and npm to fetch official packages and components from the npm registry. These operations target well-known, trusted developer services (shadcn, Tailwind Labs) and are standard for the intended technical workflow.\n- [SAFE]: The provided design philosophy in references/canvas-design-system.md contains specific stylistic instructions and quality benchmarks for visual generation. These are purely instructional and do not attempt to override agent safety filters or system instructions.\n- [SAFE]: The binary .coverage file is a standard SQLite database produced by the coverage.py testing tool. While it contains metadata about the author's local file paths, it does not include sensitive credentials, secrets, or executable payloads.

ui-styling