managing-python-dependencies

First detect the project's dependency manager by checking for files like uv.lock, pyproject.toml, Pipfile, or requirements.txt. Then use the matching command, such as `uv add`, `poetry add`, or `.venv/bin/pip install`, instead of a global pip install.

Check the workspace for signal files in priority order: uv.lock or [tool.uv] in pyproject.toml indicates uv, [tool.poetry] indicates Poetry, Pipfile indicates Pipenv, environment.yml indicates Conda, and requirements.txt alone indicates venv + pip.

Default to venv + pip with a requirements.txt file. Create an isolated environment with `python3 -m venv .venv`, install packages via `.venv/bin/pip install <package>`, and preserve state with `.venv/bin/pip freeze > requirements.txt`.

Global pip installs bypass the project's established dependency manager, pollute the system environment, and break reproducibility. The Skill prohibits global installs and overriding an existing manager with a different tool.

No. When a project already uses Poetry, indicated by [tool.poetry] in pyproject.toml, you must continue using `poetry add` and `poetry install`. Overriding the established manager with a different tool is prohibited.