The Agent Skills Directory

Command Execution: The skill uses subprocess.run to invoke the ruff formatting utility on generated source code. This is implemented using a static list of arguments, ensuring the command is localized to the formatting task and preventing shell injection.
File System Operations: The tool writes Python test files (test_runnability.py) and configuration files (conftest.py) to the targeted recipe's tests/ directory. It includes safety mechanisms such as requiring an explicit --overwrite flag to clobber existing files and a dry-run mode to prevent unintended changes.
Indirect Prompt Injection Surface: The skill processes user-provided Python code and pyproject.toml files to determine necessary test configurations.
Ingestion points: .py files and pyproject.toml files within the user-specified recipe directory.
Boundary markers: The skill instructions mandate a dry-run first and require the agent to present a preview of the generated code to the user before writing to disk.
Capability inventory: File writing (Path.write_text) and subprocess execution (ruff formatter).
Sanitization: The script uses the ast (Abstract Syntax Tree) module for structured parsing of Python files, which is more resilient than raw text processing for identifying code patterns.

generate-python-runnability-test