The Agent Skills Directory

EXTERNAL_DOWNLOADS (HIGH): The skill uses pnpm dlx shadcn@latest init to download and execute code from the npm registry. As the 'shadcn' package source is not within the specifically defined trusted organizations, this is treated as an untrusted external download.
REMOTE_CODE_EXECUTION (HIGH): Using pnpm dlx (similar to npx) to run the shadcn initialization script constitutes downloading and then executing code at runtime. This provides a direct path for remote code execution if the package or its registry entry were compromised.
COMMAND_EXECUTION (HIGH): The skill requires the agent to execute shell commands to perform initialization and framework detection, granting the skill the ability to run arbitrary subprocesses.
PROMPT_INJECTION (HIGH): The skill exhibits a high-tier Indirect Prompt Injection surface. It reads external project files and then performs write/execute operations based on that content.
Ingestion points: Reads existing tsconfig.json, jsconfig.json, next.config.js, vite.config.ts, and project CSS files.
Boundary markers: Absent; there are no delimiters or instructions to ignore malicious content within these project files.
Capability inventory: Shell execution (pnpm dlx) and extensive file system write/modification capabilities across the project directory.
Sanitization: None; the skill parses and trusts the structure of the existing configuration files to determine its next actions.

shadcn-setup