Python Uv
Standardize Python deps on uv instead of pip or Poetry when scaffolding or maintaining solo projects and agent scripts.
Overview
python-uv is an agent skill for the Build phase that enforces fast, reproducible Python dependency management exclusively through uv.
Install
npx skills add https://github.com/mindrally/skills --skill python-uvWhat is this skill?
- Mandates uv for add, remove, sync—never pip, pip-tools, or Poetry for dependency management
- Supports PEP 723-style inline script metadata blocks for standalone .py files
- CLI flags to add/remove packages and sync per script (--script script.py)
- Runs scripts via uv run with dependencies resolved automatically
- Five explicit principles including locked, synchronized environments across machines
- 5 key principles for uv-only package operations
Adoption & trust: 584 installs on skills.sh; 133 GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
Your Python project mixes pip habits with modern packaging and you cannot trust that every environment installs the same locked dependency set.
Who is it for?
Solo builders starting or maintaining Python repos, scripts, and agent utilities who want Astral uv as the single source of truth for deps.
Skip if: Teams standardized on Poetry-only workflows who will not adopt uv, or non-Python stacks where this skill adds no value.
When should I use this skill?
User is adding Python packages, syncing environments, or running scripts and might default to pip, pip-tools, or Poetry.
What do I get? / Deliverables
The agent uses uv add, remove, sync, and uv run consistently—including inline script metadata—so dependencies stay locked and synchronized without falling back to pip or Poetry.
- uv-managed dependency changes
- synchronized lockfile and environment
- script metadata or pyproject updates
Recommended Skills
Journey fit
Dependency install, lock, and sync happen continuously while you implement Python backends, CLIs, and one-off automation scripts. Backend work is where pyproject.toml, virtualenvs, and reproducible installs matter most before ship and operate.
How it compares
Use instead of ad-hoc pip install instructions when you want one resolver and lockfile workflow aligned with modern uv projects.
Common Questions / FAQ
Who is python-uv for?
Indie developers and agent users writing Python backends, CLIs, and standalone scripts who need consistent, fast dependency installs with uv.
When should I use python-uv?
During Build when adding packages, syncing after git pull, configuring script-local deps, or migrating off pip/requirements.txt before you ship or run automation in CI.
Is python-uv safe to install?
Review the Security Audits panel on this Prism page for the upstream skill package; the skill itself instructs package management commands you should run only in repos you trust.
SKILL.md
READMESKILL.md - Python Uv
# Python Package Management with uv You are an expert in Python development with uv package management. ## Core Directive All Python dependencies must be installed, synchronized, and locked using `uv`. Never use `pip`, `pip-tools`, or `poetry` directly for dependency management. ## Dependency Management Commands For standard projects: ```bash uv add <package> uv remove <package> uv sync ``` ## Script Management Execute scripts with proper dependency handling: ```bash uv run script.py ``` ### Manual Inline Metadata Configuration Scripts can specify dependencies via comment blocks: ```python # /// script # requires-python = ">=3.12" # dependencies = [ # "torch", # "torchvision", # "opencv-python", # "numpy", # "matplotlib", # "Pillow", # "timm", # ] # /// print("some python code") ``` ### CLI-Based Script Dependencies ```bash uv add package-name --script script.py uv remove package-name --script script.py uv sync --script script.py ``` ## Key Principles 1. Always use `uv` for all package operations 2. Prefer inline script metadata for standalone scripts 3. Use `uv run` to execute scripts with their dependencies 4. Keep dependencies locked and synchronized across environments 5. Never fall back to pip or other package managers