The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The scripts/init-artifact.sh script is vulnerable to shell command injection. The $PROJECT_NAME variable is unsafely interpolated into a sed command line without sanitization or proper quoting (e.g., '$PROJECT_NAME'). This allows a malicious user or input to execute arbitrary shell commands on the host system.
[COMMAND_EXECUTION] (MEDIUM): The initialization script performs a global installation of the pnpm package manager using npm install -g pnpm if it is not detected. Global installations can modify system-wide configurations and typically require elevated permissions.
[REMOTE_CODE_EXECUTION] (HIGH): The skill utilizes pnpm create vite and pnpm install, which download and execute code from the npm registry at runtime. This introduces a significant reliance on the security of the external package ecosystem.
[EXTERNAL_DOWNLOADS] (LOW): The skill downloads and installs over 50 external Node.js packages during the project setup and bundling phases (including Vite, Tailwind CSS, and various Radix UI components).
[PROMPT_INJECTION] (LOW): The SKILL.md file contains instructions labeled 'VERY IMPORTANT' that attempt to override the model's creative style (e.g., avoiding 'AI slop'), which is a form of behavioral instruction override.

artifacts-builder