python-conventions（Python conventions 测试）

Python Conventions

Apply these conventions when Python work is the primary workstream.

Dispatch

Reference File Index

Operator Contract

Active

1. Apply this skill only when Python files, Python tooling, or pyproject.toml are the primary surface of the task.
2. Read references/redirection-boundaries.md when Python appears alongside shell, JS/TS, or other dominant workstreams.
3. Enforce the hard requirements in references/tooling-contract.md for package management, command execution, linting, type checking, and tests.
4. Check references/exceptions.md before recommending any legacy or constrained-environment deviation.
5. Read references/library-preferences.md only when choosing libraries for new Python work.

Empty / Help

1. Summarize the hard requirements: uv, uv run, uv add, ty, ruff, pytest, and pyproject.toml.
2. Show the difference between hard requirements and guided preferences.
3. Point to the exact reference files for tooling, exceptions, libraries, testing, performance, and mixed-language routing.

check

1. Verify tooling compliance only; do not widen into full implementation advice unless the user asks.
2. Report whether the project uses uv, uv run, ty, ruff, pytest, and pyproject.toml in the expected ways.
3. Flag legacy or exception-path deviations and require the reason to match references/exceptions.md.
4. Reject recommendations that replace repo-required tooling with mypy, pip install, or bare python.

Hard Requirements

• Package manager: use uv for all Python package operations
• Dependencies: use uv add, uv add --group dev, uv remove, and uv lock --upgrade-package <pkg> as appropriate
• Reproducible installs: use uv sync --locked when validating an existing lockfile workflow
• Project config: keep Python project configuration in pyproject.toml
• Type checking: use uv run ty check, not mypy
• Linting: use uv run ruff check and uv run ruff format
• Testing: use uv run pytest
• Task running: use uv run <command> for Python command execution

Virtual Environments

• Let uv manage virtual environments automatically
• Never manually create or activate .venv directories
• Use uv run to execute within the project environment

Guided Preferences

Guided library preferences apply only when starting new Python work and no stronger local constraint already exists. Read references/library-preferences.md before recommending replacements for an established stack.

Project Structure

• Use pyproject.toml for all project metadata and dependencies
• Place source code in a package directory matching the project name
• Use uv workspace members for monorepo sub-packages
• Run the required lint, format, type-check, and test sequence from references/tooling-contract.md before considering Python work complete

Performance Conventions

1. Profile before optimizing.
2. Use references/performance-tips.md for quick Python profiling and optimization patterns.
3. Route broad profiling, regression analysis, or performance investigation to performance-profiler.

Testing Conventions

1. Use pytest for Python tests and configure project test defaults in pyproject.toml.
2. Use references/testing-patterns.md for fixtures, markers, tmp_path, monkeypatch, parametrization, and coverage policy.
3. Route test strategy, suite design, fixture architecture, or cross-language test plans to test-architect.

Validation Contract

Before declaring changes to this skill complete, run:

1. uv run wagents validate
2. uv run wagents eval validate
3. uv run python path/to/audit.py skills/python-conventions/ --format json
4. uv run wagents package python-conventions --dry-run
5. git diff --check

After changing skill definitions, public descriptions, reference files, or eval behavior, invoke docs-steward if available and then run uv run wagents readme --check.

Critical Rules

1. Reject Python setup advice that uses pip install, uv pip install, or bare python when the task is not explicitly on an approved exception path.
2. Require uv add for runtime dependencies and uv add --group dev for development-only dependencies unless references/exceptions.md justifies a legacy or constrained-environment deviation.
3. Require uv run ty check for type checking; do not recommend mypy or bare ty check as the default path in this repo.
4. Treat uv run ruff check, uv run ruff format, uv run ty check, and uv run pytest as the default completion gate for Python changes unless an exception is documented.
5. Do not edit uv.lock by hand; use uv lock, uv sync, or dependency commands.
6. Do not present guided library preferences as mandatory replacements for an already-established local stack.
7. Read references/exceptions.md before approving a legacy toolchain, alternate environment manager, or alternate library path.
8. Redirect mixed-language or non-Python-primary work through references/redirection-boundaries.md instead of force-fitting this skill onto the whole task.

Canonical terms (use these exactly):

• uv -- the required package manager and task runner
• ty -- the required type checker (not mypy)
• uv run ty check -- the required type-check command
• ruff -- the required linter and formatter
• pyproject.toml -- the single source of project configuration
• uv run -- prefix for all Python command execution
• uv sync --locked -- default reproducible install check for lockfile-backed workflows

Scaling Strategy

• Incidental Python file in a broader non-Python task: enforce only the hard requirements that touch the Python-owned surface, then route mixed-workflow questions through references/redirection-boundaries.md.
• Python-primary feature or refactor work: apply the full operator contract, including tooling, testing, and guided preferences where relevant.
• Repo-wide Python tooling or migration work: use check, references/tooling-contract.md, and references/exceptions.md to separate hard violations from documented transition paths.

Progressive Disclosure

• Do not load every reference by default.
• Read references/tooling-contract.md first for command-sequence questions.
• Read references/redirection-boundaries.md when shell, JS/TS, CI, or framework-specific work is mixed into the request.
• Read references/exceptions.md only when the task appears to require legacy, corporate, or constrained-environment exceptions.
• Read references/library-preferences.md only when selecting libraries for new Python work.
• Read performance or testing references only when the active task actually touches those areas.

Scope Boundaries

IS for: Python tooling conventions, command selection, dependency-management rules, type/lint/test gates, and exception-aware repo guidance.

NOT for: JS/TS conventions, shell conventions, CI pipeline design, profiling investigations, test architecture, or framework/domain-specific implementation strategy.