Python Productive

Instructions

Help teams ship Python 3.11+ with uv (installer + resolver + lock), Pydantic v2, and strict typing. Read upstream docs for the versions the user pins (uv docs, Pydantic migration guide, Ruff rules) before suggesting APIs - do not invent removed Pydantic v1 APIs.

When to Use

• Use for uv, Pydantic v2, typing, and productive Python tooling.
• Prefer Next/React skills when the task is frontend.
• Prefer secure-dependencies mindset for lockfiles and audits.

Mode A - greenfield or refresh

1. Prefer pyproject.toml with [project] metadata, dependencies / optional [dependency-groups] for dev tools.
2. Use uv lock / uv sync patterns; document UV_PROJECT_ENVIRONMENT only if user needs it.
3. Pydantic v2: model_config, field_validator, computed_field; avoid @validator / class Config v1 patterns.
4. Ruff for lint + import sort; align with formatter (Ruff format or Black - pick one per repo).
5. Types: basedpyright or mypy - start from strict subset the repo tolerates.
6. Tests: pytest layout (tests/ or src layout); one example test command.

Mode B - audit

1. List deprecated patterns (Pydantic v1, setuptools-only layout without pyproject).
2. Flag slow hot paths; suggest profile first, then optional Rust extension (maturin) or Cython only when user owns build complexity.

Native speed-ups (when asked)

• Small inner loops: C extension or maturin crate behind a thin Python API - document build (uv build) and platform wheels.
• Validation-heavy paths: Pydantic core is already Rust-backed; prefer model design over custom C unless profiling proves need.

Outcomes

• Actionable pyproject + tool config diffs or audit tables.

Output Rules

State Mode A or B. Show exact uv / pytest commands matching the repo.

Scope and boundaries

• In scope: application Python, tooling, typing, packaging with uv.
• Out of scope: rewriting large scientific stacks without profiling, publishing to PyPI without user keys.

Safety

• Do not print .env or API keys; use Pydantic SecretStr patterns when showing examples.
• Match repo’s existing Python version floor.

Troubleshooting

• Lock conflicts: regenerate lock with uv lock after bounds change.
• Pydantic errors: link field name + error type; avoid blanket model_rebuild() everywhere.

Related skills

• testing-agent - if testing Python from a JS monorepo host
• secure-dependencies - lockfiles and supply chain mindset

GitHub: https://github.com/bh611627/skillcodex/tree/main/skills/python-productive/SKILL.md
npm: https://www.npmjs.com/package/@skillcodex/skills