write-code

name	write-code
description	Write or review code in this repo. Use when writing new code, reviewing existing code, or editing code.
argument-hint	[source code]

Writing Code

Use this skill when writing, reviewing, or editing code.

Purpose

The goal of this document is to provide guidance when coding in this repository. All code should be pythonic and easy to read.

Python version: >=3.12

Before every commit

• Ensure all code has type annotations
• Add Google-style docstrings (NO types, NO leading articles)
• Run checks: prek -a
• Prek hooks will run automatically and must pass

Code conventions

Docstrings

See skill write-docstrings

Type annotations

• All functions must have complete type annotations
• Import types from typing only when necessary (prefer built-ins)
• Use modern syntax: list[str], dict[str, int] (not List[str], Dict[str, int])
• Do not use a bare dict, always annotate the type of the dict (e.g. dict[str, float])
• Use Any in dict annotations only if absolutely necessary
• Use | for union types: str | None
• Avoid from __future__ import annotations
• Use modern numpy type hints where useful, and use TypeAlias to make code more readable, e.g.
  • type Matrix[T: np.generic] = np.ndarray[tuple[int, int], np.dtype[T]]

Code formatting

Via ruff

• Line length: 100
• Indentation: 4 spaces (no tabs except Makefiles)

Naming conventions

• Functions/methods: snake_case
• Variables: snake_case
• Constants: UPPER_SNAKE_CASE
• Classes: PascalCase
• Modules: snake_case
• Private attributes/methods: _leading_underscore

Imports

• Absolute imports preferred
• Group imports: standard library, third-party, local
• No wildcard imports (from module import *) except in __init__.py
• Import sorting handled by ruff (isort)
• Do not use import statements within a function unless absolutely necessary

Error handling

• Prefer specific exceptions; define custom exceptions for domain errors (e.g. SpamTypeError(ValueError), SpamEatingError(RuntimeError))
• Avoid bare except: clauses

General style

• Use f-strings for string formatting
• Prefer list/dict comprehensions over loops when appropriate
• Use pathlib.Path for file operations instead of os.path
• Use dataclasses and prefer the settings slots=True and frozen=True.
• Don't use .0 to indicate floats
• Prefer strict=True in zip and itertools.batched
• Use a guard case in all match/case statements (i.e. case _:)

Essential commands

Code quality tools (ruff, ty, pytest) are configured per-package in pyproject.toml.

# Setup
pixi install                    # Install dependencies
prek install                    # Install git hooks

# Code quality
pixi run fmt                    # Format code
pixi run lint                   # Lint code
pixi run types                  # Type check
prek -a                         # Run all prek hooks
prek run <hook-id>              # Run specific hook

# Testing
pixi run test                   # Run tests
pixi run test --cov             # Run tests with coverage

# Package management
pixi add <package>              # Add dependency
pixi add --dev <package>        # Add dev dependency
pixi install                    # Install/sync dependencies

Direnv integration

Direnv auto-activates the pixi dev environment when entering the project directory. Run direnv allow . if you see permission errors.

Testing

• Use pytest; tests live in tests/; doctests are auto-discovered in source
• Use standard pytest format (e.g. don't use classes to hold tests)
• For float comparisons, use approx or assert_almost_equal (imported as aae), prefer default thresholds

See the skill write-tests for more detail, but only if actively writing tests

CI/CD

CI runs pixi run fmt, pixi run lint, pixi run types, and pixi run test per-package via GitHub Actions.

Git development guidelines

When working on a new feature, bugfix, etc. If Claude is already working in this directory, offer to start a new worktree. Create a new branch for all new work (avoid developing on master)

e.g. git worktree add ../{package_name}-{feature_name} && cd ../{package_name}-{feature_name} # optional git switch -c feat/jalapeno_spam

Break up work into manageable commits, keeping track of what changed in each commit. When the work is complete, create a descriptive PR, breaking down what was added, changed, and removed.

e.g. gh pr create --title "Adds Jalapeno Spam" --body "Spam is a tasty processed meat ..."

If a worktree was created, offer to clean it up.

Commit guidelines

Format: conventional commits recommended

• feat: - new features
• fix: - bug fixes
• docs: - documentation changes
• test: - test changes
• refactor: - code refactoring
• chore: - maintenance tasks

Never include Claude as a co-author on commits.

Example

git commit -m "feat: add Jalapeno Spam

- Adds Jalapeno Spam
- Tests that it is spicy"

Documentation

Use skill write-docstrings

Additional resources

• pixi documentation: https://pixi.sh
• ruff documentation: https://docs.astral.sh/ruff
• ty documentation: https://github.com/astral-sh/ty
• pytest documentation: https://docs.pytest.org
• prek documentation: https://prek.j178.dev
• Google docstring style: https://google.github.io/styleguide/pyguide.html#38-comments-and-docstrings

Troubleshooting

Prek hook failures

Formatting issues:

• Usually auto-fixed by ruff
• Re-stage files: git add .
• Try committing again

Lint issues:

• Read error message for specific rule
• Fix; if # noqa: <rule> if absolutely necessary, ask before adding

Type issues:

• Add missing type annotations
• Fix type mismatches
• Use pixi run types to verify locally

Test failures:

• Fix failing tests or code
• Run pixi run test -v for detailed output
• Run specific test: pixi run test tests/test_file.py::test_name

Miscellaneous

Do not add new packages without explicitly asking first. Never add ignores to the pyproject.toml formatting, linting, or type checking without explicitly asking