Verified Agent Skill record

Python Env

A guide for setting up and installing into Python environments, requiring that the environment live with the project: created inside the project directory as `.venv`, or managed by uv, never in a scratch location such as /tmp and never installed into the system interpreter.

Tasks

Create a project-local virtual environment (.venv) at the project root for a Python project
Set up a greenfield Python project with uv, including uv venv and installing the package with uv pip install -e . or uv sync when a lockfile exists
Write run instructions that work from a fresh shell in the project directory using the project environment
Choose the correct response when pip install fails with externally-managed-environment instead of working around it
Detect and reuse an existing environment or declared tool (.venv, uv.lock, Poetry, Pipenv, conda) before creating a new one

Inputs

A Python project directory
An existing environment or dependency tool declaration (e.g., .venv, uv.lock, Poetry, Pipenv, conda), if present
A lockfile, if present

Outputs

A project-local virtual environment (`.venv` inside the project directory)
Run instructions written against the project environment

Limitations and checks

Not user-invocable per the source frontmatter
The source does not document platform support beyond noting that Homebrew, Debian, and Ubuntu mark the system interpreter as externally managed
Confirm the environment is inside the project directory (`.venv` at the project root), not in /tmp or another scratch location
Confirm pip install was not run against the system interpreter with --break-system-packages, PIP_BREAK_SYSTEM_PACKAGES, or a deleted EXTERNALLY-MANAGED marker
Confirm run instructions reference the project environment from a fresh shell and do not use absolute paths to an environment outside the project
Confirm no second environment or tool was introduced when a .venv, uv.lock, Poetry, Pipenv, or conda setup already exists