← ClaudeAtlas

uv-project-workflowlisted

Manage uv dependencies, locks, environments and CI; not unrequested migrations.
ValentinNikolaev/agent-skills · ★ 1 · AI & Automation · score 74
Install: claude install-skill ValentinNikolaev/agent-skills
# uv Project Workflow Use uv when the repository already chooses it or the user requests its adoption. Make the requested environment/dependency change reproducible while preserving existing work, supported runtimes and source policy. ## Route precisely - Use for explicit uv project setup, dependency/lockfile maintenance, environment diagnosis, CI configuration or an authorized migration to uv. - Do not choose uv merely because code is Python. Preserve pip, Poetry, Conda or another established workflow unless changing it is in scope. - Use `code-review` for an existing manifest/lockfile diff without edits, `vet-dependency` for whether to adopt a library, and `deprecation-and-migration` for a live application/API/schema migration. Implementing a FastAPI endpoint is not an environment-management request. - Distinguish uv's project workflow from its pip-compatible interface, tool environments and standalone scripts; do not mix their sources of truth. ## Inspect before mutation 1. Read repository instructions and working-tree state. Identify the owning project/workspace and any nested independent project. Preserve dirty manifests, lockfiles and parallel work. 2. Read `pyproject.toml`, `uv.lock`, `.python-version`, workspace/source/index configuration, existing requirements/constraints, CI and runtime targets. Resolve competing lockfiles or conflicting tool selection before writing. 3. Record the installed uv and Python versions, requested outcome, direct dependencies, groups