The Agent Skills Directory

[SAFE]: The project initialization workflow follows security best practices by implementing a multi-stage process where a read-only preflight check and a blueprint plan must be approved by the user before any mutations are performed.
[SAFE]: The skill contains active credential protection mechanisms. The scripts catalog_utils.py and validate_plan.py identify 'secret_read' command patterns (such as catting .env files or extracting tokens via CLI tools) and explicitly block them from being included in any execution plan.
[COMMAND_EXECUTION]: The internal Python scripts utilize subprocess calls to perform local environment audits, repository status checks, and template processing. These operations are appropriately scoped to project metadata and are protected by the skill's planning and approval gates.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of common development dependencies and scaffolds using well-known package managers such as npx, pnpm, and uv. These operations target established registries (NPM, PyPI) and are only performed after explicit user approval of the project blueprint.

new-project