The Agent Skills Directory

Command Execution: The skill utilizes docker and uv commands via the subprocess module to build container images, run verification containers, and manage Python project dependencies. These operations are core to the skill's purpose of containerizing and verifying agent recipes.
External Downloads: During operation, the skill may download standard Python packages (e.g., tomlkit, ruamel.yaml, packaging, httpx) and base Docker images (e.g., python:3.12-slim). These downloads originate from well-known registries and are necessary for the tool's functionality.
Execution of Project Code: The skill performs a 'boot check' and 'container verification' which involves executing code from the Python recipe being processed. Specifically, it imports the application object from the recipe's package into a test client and runs the generated container. While this involves running local project code, the skill implements a run_allowlist and requires human confirmation before executing containers, which are standard safety measures for development tools.
Local File Operations: The skill reads and writes several project files, including pyproject.toml, agent.py, manifest.yaml, and a new Dockerfile. These modifications are performed to align the project with deployment standards and are limited to the target recipe directory.

make-python-recipe-deployable