coding-style
Apply Andrii's Python and TypeScript coding standards when implementing, simplifying, or reviewing code, APIs, packages, and tests. Use for explicit types, module boundaries, public exports, readable control flow, and useful comments.
- License 许可
- MIT
- 所属仓库
- yasik/skills
- 最近更新
- 2026-11-08
- Stars
- ★ 3
- Forks
- 1
01 前言 Front Matter
- name
- coding-style
- description
- Apply Andrii's Python and TypeScript coding standards when implementing, simplifying, or reviewing code, APIs, packages, and tests. Use for explicit types, module boundaries, public exports, readable control flow, and useful comments.
02 资源文件 15
- 参考文档references/SOURCE-LICENSE1.0 KB
- 参考文档references/source.md1.2 KB
- 参考文档references/standards/documentation.md1.1 KB
- 参考文档references/standards/python/comments.md9.4 KB
- 参考文档references/standards/python/imports.md2.2 KB
- 参考文档references/standards/python/modules.md2.5 KB
- 参考文档references/standards/python/python.md5.8 KB
- 参考文档references/standards/python/README.md892 B
- 参考文档references/standards/python/testing-and-prs.md1.7 KB
- 参考文档references/standards/python/tool-design.md4.1 KB
- 参考文档references/standards/README.md3.0 KB
- 参考文档references/standards/typescript/conventions.md8.3 KB
- 参考文档references/standards/typescript/README.md677 B
- 参考文档references/standards/typescript/testing-and-tooling.md3.3 KB
- 参考文档references/standards/workspace-packages.md1.7 KB
03 SKILL.md 正文
Coding style
Use the bundled standards as the coding baseline. Read the target repository's instructions, manifests, and nearby code first. Apply the relevant rules to the requested change; do not use this skill to start an unrelated repository-wide rewrite.
Read the relevant references
Start with the shared rules, then select:
| Work | Reference |
|---|---|
| Python | Python index, then its applicable pages |
| TypeScript | TypeScript conventions |
| New package or export | Workspace packages and the language's module/export rules |
| Tests and tooling | Python or TypeScript |
| Agent tools | Tool contracts, when the target has equivalent tool boundaries |
| Documentation | Documentation rules, with the adaptation limits below |
Read adaptation notes before choosing commands, dependencies, or package structure. All reference paths are relative to this skill; no installed harness API is required.
Apply the standards
- Keep domain, command, response, event, and callback boundaries explicitly typed. Parse external data at the boundary. Reuse the owning public type instead of creating a second vocabulary.
- Keep public exports explicit and internal modules private. Put shared code at its actual ownership boundary. Justify any new helper, abstraction, dependency, or compatibility fallback by a current need.
- Preserve useful extension seams, invariants, and error contracts. Simpler code must not silently become a weaker contract.
- Separate logical blocks with blank lines. Use early guards where they make the happy path clear. Explain constraints and non-obvious decisions in comments; remove comments that merely narrate code.
- Apply the detailed language rules, including Python import/annotation policy and TypeScript strictness, named exports, and public JSDoc. Keep formatter and type-check configuration owned by the target repository.
- Run applicable existing checks. Inspect the final diff for readability and accidental export, behavior, or scope changes; formatting alone is not a style review.
Use the unit-test-coverage skill when deciding what tests earn their place.
If it is unavailable, require each test to protect a meaningful public behavior
or failure mode rather than private wiring or a coverage number.
Adaptation boundaries
The references define coding preferences, with illustrative package and tool contracts. Use the target repository's declared runtime, dependencies, formatter, and build commands. Do not introduce a tool framework or change the package manager to satisfy an example. Target instructions and explicit user requirements take precedence. Raise only unresolved conflicts that materially affect the work.
Report concrete violations fixed or remaining, relevant files, and checks run. Do not produce a generic checklist of rules that had no effect on the change.
04 同仓库更多 Skills
terminal-report
Produce beautiful, structured terminal output for a CLI or analysis tool — ANSI colors, aligned data tables, inline bar charts, diverging/centered bars, tagged progress logs, sparklines, and section headers. Use this skill whenever you are building a command-line tool, analysis script, report generator, data dashboard, or status/progress display that prints to a terminal, OR whenever the user wants output to look "polished", "colored", "styled", "like a dashboard", or shows a screenshot of a pretty terminal and asks how to make one. Applies even when the user just says "print the results nicely" or "make a little CLI for this" without naming colors or charts explicitly.
technical-prose
Write and revise technical explanations, design docs, reviews, reports, and progress updates in clear, connected prose. Use when the reader needs to understand a mechanism or decision, or when a draft feels dense, choppy, jargon-heavy, or like compressed implementation notes. Preserve technical meaning while making the reasoning easy to follow.
teach
Teach a complex topic as one complete, self-paced written lesson — chaptered sections the learner reads top to bottom, a deep-dive that slows down on the single hardest concept and breaks it into steps with worked examples, diagrams only where they actually clarify, and real citations for going deeper. Works from an explicit topic ("teach me how Raft consensus works") or from one or more links/papers/docs the user wants to be brought up to speed on — it fetches them, falling back to a browser when a page is paywalled, JavaScript-rendered, or only partially retrievable. Use this whenever the user wants to *understand* something rather than just get an answer: "teach me X", "explain this <link>", "help me understand <topic>", "I want to learn <thing>", "break this down for me", "ELI5 but properly / but deep", or when they drop a URL and ask to be walked through it like a class. Distinct from a quick factual lookup or a terse summary — reach for this when the goal is durable understanding, not a TL;DR.
mermaid-render
Render a Mermaid diagram as a beautiful, optionally animated SVG (and PNG) in a calm editorial style, instead of stock Mermaid output. Supports flowchart, stateDiagram, and sequenceDiagram. Use whenever the user wants to render, draw, beautify, restyle, or "make nice" a Mermaid / flowchart / sequence / state diagram, add animated data-flow to a diagram, or turn a described process or architecture into a polished diagram for a doc, README, or slide. All styling — colors, fonts, shapes, borders, animation — is configurable in one theme file.
dev-docs
Write, rewrite, or review developer documentation to a complete, packaged technical-writing style guide: tone, grammar, punctuation, formatting, procedures, code samples, placeholders, UI text, linking, naming, and a word list of about 700 terms, plus an index of every rule. Use whenever the user asks for developer docs, technical docs, a README, API reference, quickstart, how-to, tutorial, concept page, release notes, docstrings or reference comments, or a docs style review, or wants docs that read like a major platform's documentation. Also use to check a draft against the style guide or to look up a term in the word list.
unit-test-coverage
Select, write, and review meaningful tests of public functions and interfaces. Use when adding behavior, fixing a regression, reviewing test quality, or diagnosing failing tests. Every test must protect a critical behavior or flow; never add tests only to increase coverage.