The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/init-artifact.sh is vulnerable to shell command injection. The $PROJECT_NAME variable is interpolated into a sed command string using double quotes (e.g., '$PROJECT_NAME'). In shell scripts, double-quoted variables are still subject to command substitution, meaning an attacker can execute arbitrary code by providing a project name containing backticks or $(...) syntax.
[EXTERNAL_DOWNLOADS]: The scripts/init-artifact.sh and scripts/bundle-artifact.sh scripts download and install a large number of Node.js packages from the public npm registry to bootstrap the React environment and bundle artifacts.
[COMMAND_EXECUTION]: The skill performs several high-impact system operations, including the global installation of the pnpm package manager (npm install -g pnpm) if it is not detected, and uses the sed utility to modify files in place.
[DYNAMIC_EXECUTION]: The initialization process executes inline JavaScript via node -e to programmatically modify project configuration files like tsconfig.json.
[METADATA_POISONING]: The LICENSE.txt file explicitly attributes copyright to 'Anthropic, PBC', while the skill's metadata lists the author as 'siraxuth'. This discrepancy is misleading regarding the skill's true origin and safety.

web-artifacts-builder