The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes two shell scripts (init-artifact.sh and bundle-artifact.sh) that perform extensive command-line operations.
bundle-artifact.sh: Executes pnpm to install dependencies and run build tools (parcel, html-inline), and uses du to check file sizes.
init-artifact.sh: Uses sed for file modification, node -e for programmatic JSON manipulation of tsconfig.json, tar for archive extraction, and manages project creation via pnpm create.
[EXTERNAL_DOWNLOADS]: The scripts trigger the installation of a large number of Node.js packages from the public npm registry.
bundle-artifact.sh installs parcel, @parcel/config-default, parcel-resolver-tspaths, and html-inline.
init-artifact.sh installs over 50 packages including Vite, Tailwind CSS, and numerous @radix-ui primitives.
[PRIVILEGE_ESCALATION]: The init-artifact.sh script attempts a global installation of a package manager if it is missing.
npm install -g pnpm: This operation typically requires administrative/root privileges (sudo) depending on the user's Node.js environment configuration.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface where user-provided design intents are interpolated into code generation tasks.
Ingestion points: User design requests (e.g., "create a dashboard") in SKILL.md.
Boundary markers: Absent; there are no specific delimiters or instructions to ignore malicious embedded commands in the user's design requirements.
Capability inventory: The skill can write files and execute shell scripts through the provided .sh files.
Sanitization: Absent; the skill does not explicitly validate or sanitize the generated UI code for potential script injections (e.g., malicious useEffect hooks).

visual-designing