--- name: python-uv-development description: Implement and troubleshoot Python projects with supported runtimes, isolated environments, uv and repository-specific checks. --- # Python and uv - Read repository instructions, manifests, lockfiles and existing commands. Preserve the project's Python range and package manager. - Never replace Ubuntu's system Python, install into its environment, or use `--break-system-packages`. Select a compatible user-managed runtime explicitly. - Keep environments project-local. For uv projects use intentional `uv add`, `uv sync` and `uv run` changes; do not create one environment for every repo. - Change manifests and generated lockfiles through the correct tooling. Personal tools must not override a project's pinned dependency versions. - Run focused formatting, lint, type and test checks for the changed behavior. Do not hide failures with broad catches or silent defaults. - Distinguish runtime, import, package and native-library errors. If a particular runtime or wheel is incomplete, repair that installation/cache rather than deleting all environment or user configuration data. For GPU Python also use `nvidia-cuda-workflows`. Framework runtime wheels and a native CUDA development toolkit are not interchangeable.