The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The script scripts/init-artifact.sh attempts to install the pnpm package manager globally using npm install -g pnpm if it is not already present in the environment. Global installations generally require elevated system privileges (e.g., sudo or Administrator access) to modify system-wide directories.
[COMMAND_EXECUTION]: The initialization script scripts/init-artifact.sh accepts a user-provided project name as a command-line argument ($1) and uses it directly within shell commands and sed operations without sanitization. For example, it is used in pnpm create vite "$PROJECT_NAME" and $SED_INPLACE 's/<title>.*<\/title>/<title>'"$PROJECT_NAME"'<\/title>/' index.html. A maliciously crafted project name could lead to command injection or unintended file modifications.
[DYNAMIC_EXECUTION]: The script scripts/init-artifact.sh utilizes node -e to execute dynamic JavaScript code snippets at runtime to parse and modify tsconfig.json and tsconfig.app.json files.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs numerous development dependencies and libraries from the NPM registry using pnpm and npm. These include established tools and libraries such as Vite, Tailwind CSS, Radix UI components, and Parcel.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data (the project name) which is then interpolated into prompts and shell script execution paths.
Ingestion points: The scripts/init-artifact.sh script accepts a project name as its first positional argument.
Boundary markers: No delimiters or instructions to ignore embedded commands are present when handling this input.
Capability inventory: The skill has capabilities for shell command execution, file system writes, and package installation via pnpm, npm, and node.
Sanitization: There is no evidence of sanitization, escaping, or validation performed on the input project name before it is used in sensitive operations.

artifacts-builder