The Agent Skills Directory

[COMMAND_EXECUTION]: The skill depends heavily on shell scripts (scripts/init-artifact.sh and scripts/bundle-artifact.sh) that perform broad operations on the filesystem, including directory creation, project scaffolding, deletion of build outputs (rm -rf dist), and programmatic configuration file management.
[PRIVILEGE_ESCALATION]: In scripts/init-artifact.sh, the skill attempts to install the pnpm package manager globally using npm install -g pnpm. This action modifies the host system environment and typically requires administrative privileges.
[EXTERNAL_DOWNLOADS]: The scripts install a large number of Node.js dependencies from the public npm registry. Many of these packages are not pinned to specific versions, which introduces supply chain risks through potential malicious or compromised updates.
[REMOTE_CODE_EXECUTION]: The initialization script executes pnpm create vite, which downloads and runs external scaffolding code to set up the project structure.
[DYNAMIC_EXECUTION]: The skill uses node -e to programmatically read and modify configuration files (tsconfig.json, tsconfig.app.json) at runtime. It also extracts a binary tarball archive (shadcn-components.tar.gz) containing component code into the project's source directory without integrity verification or path validation.
[METADATA_POISONING]: The skill includes a skill-report.json file that presents a simulated security audit report referencing non-existent AI models (e.g., gpt-5.5) and fake provenance metadata, which could be used to mislead users or agents regarding the skill's safety status and origin.

web-artifacts-builder