litestar-org/litestar-skills/litestar-styleguide

litestar-styleguide

This is the shared style baseline that every other skill in this plugin references. It exists so that cross-cutting rules (PEP 604 unions, async I/O, ruff + mypy + pyright + prek, test file naming, CI/CD conventions) live in exactly one place — and individual skills stay focused on their framework or tool-specific surface.

Code Style Rules

• Terse, imperative, authoritative. State the preferred choice and a one-line Reason:; never hedge ("you might want to…").
• PEP 604 unions (`T | None`) and PEP 585 generics (`list[T]`, `dict[K, V]`). Never use typing.Optional, typing.Union, typing.List, or typing.Dict.
• `from __future__ import annotations` is a library-author guardrail, not a consumer rule. Application code (handlers, services, tests) MAY use it; avoid it only in modules defining runtime-introspected types (msgspec.Struct, SQLAlchemy Mapped[...], Dishka @provide, SAQ @task).
• Keep runtime-inspected route annotations at module top-level (`# noqa: TC001, TC002`). When from __future__ import annotations and Ruff flake8-type-checking (TC) are active in Litestar controller/handler modules, keep Request, NamedDependency, FromDishka, FromPath, FromQuery, JSONBody, and request/response DTO imports at module top-level with # noqa: TC001, TC002 (or "**/controllers.py" = ["TC001", "TC002", "TC003"] in per-file-ignores), never inside if TYPE_CHECKING:. Reason: Litestar and Dishka inspect handler annotations at runtime during app startup and OpenAPI generation.
• Async all I/O, Google-style docstrings, function-based pytest tests (`@pytest.mark.anyio`).
• Litestar first-party stack bias with match-your-stack flexibility. Prefer msgspec, advanced-alchemy / sqlspec, litestar-granian, litestar-saq / litestar-queues when starting fresh, and match the user's existing stack when present.

Quick Reference

How sibling skills consume this

Every SKILL.md in this plugin has a ## Shared Styleguide Baseline section near the bottom. That section links to a subset of these references — only the ones that apply to the skill's language / framework mix. For example:

• skills/litestar/SKILL.md links to general.md + python.md + litestar.md
• skills/litestar-vite/SKILL.md links to general.md + typescript.md + litestar.md
• skills/litestar-testing/SKILL.md links to general.md + testing.md + python.md + litestar.md

The sibling skill extends the baseline with its own tool-specific Code Style Rules, Quick Reference, Guardrails, and Validation — but it does not duplicate the baseline. If a convention is generic (type hints, naming, imports), it belongs here.

When to update this skill

• A rule becomes contentious across two or more sibling skills → pull it into the right baseline reference file here.
• A new language lands (Rust, Mojo, etc.) → add a new references/<lang>.md and link from skills that use it.
• A tool is swapped out (e.g., ruff replaces flake8 + black, prek replaces pre-commit CLI) → update python.md / ci-cd.md once; all sibling skills inherit it.

<workflow>

Workflow — consuming this baseline

1. Open the sibling skill you are editing (skills/<name>/SKILL.md).
2. Look at its ## Shared Styleguide Baseline section — it already lists a subset of the references here.
3. When adding a rule to the sibling, ask: is it generic (language/tooling) or framework-specific? Generic → land it in the right file under references/ here. Specific → keep it in the sibling.
4. Cross-link bidirectionally if a rule here is amplified in the sibling.

</workflow>

<guardrails>

Guardrails

• No duplication across skills. A rule lives in exactly one file; sibling skills link to it.
• No folklore. Every rule has a one-line justification (perf, runtime introspection, OpenAPI alignment, etc.). Delete rules you cannot justify.
• Terse and imperative. Bullets are ≤ 2 sentences. If a topic needs more, split it into its own reference file.
• Examples are minimal and copy-pasteable. No pseudo-code; no multi-hundred-line fixtures.

</guardrails>

<validation>

Validation Checkpoint

• [ ] Every sibling skill's ## Shared Styleguide Baseline section resolves to files that exist under references/
• [ ] No rule is duplicated between two reference files (check via grep when editing)
• [ ] Each "never do X" rule has a one-line Reason: explanation
• [ ] New language support lands as a single new references/<lang>.md — not scattered into sibling skills

</validation>

<example>

Example — adding a new rule

A reviewer finds that two sibling skills independently wrote "use ruff format not black". Instead of leaving duplicates, pull the rule into references/python.md:

-**Use `ruff format`, never `black`.** Reason: ruff is the single toolchain for
  lint + format; running two formatters produces style drift.

Then in each sibling's SKILL.md, replace the duplicate with a pointer:

## Shared Styleguide Baseline
 
-[Python](../litestar-styleguide/references/python.md)

</example>

References Index

• General Principles
• Python
• Litestar
• TypeScript
• Testing
• CI/CD
• Canonical Reference Apps
• Google Developer Knowledge MCP

Official References

• <https://peps.python.org/pep-0604/> — PEP 604 union syntax
• <https://docs.astral.sh/ruff/> — ruff linter / formatter
• <https://docs.astral.sh/uv/> — uv package and environment manager
• <https://mypy.readthedocs.io/en/stable/> — mypy static type checker
• <https://microsoft.github.io/pyright/> — pyright type checker
• <https://docs.pytest.org/> — pytest

Shared Styleguide Baseline

• General Principles
• Python
• Litestar
• TypeScript
• Testing
• CI/CD