The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/init-artifact.sh script is vulnerable to shell command injection.
Evidence: The $PROJECT_NAME variable is directly concatenated into a sed command string using unsafe quoting: $SED_INPLACE 's/<title>.*<\/title>/<title>'"$PROJECT_NAME"'<\/title>/' index.html. An attacker can manipulate the input to break out of the command and execute arbitrary shell instructions.
[COMMAND_EXECUTION]: The initialization script attempts to perform global software installations which modify the host environment.
Evidence: The script executes npm install -g pnpm if the command is missing, which typically requires elevated permissions and persists beyond the skill's execution context.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs many third-party packages from public repositories.
Evidence: Both init-artifact.sh and bundle-artifact.sh install numerous npm packages (Vite, Tailwind CSS, Parcel, Radix UI, etc.) at runtime to support the project creation and bundling workflow.

artifacts-builder