AgentSkillsCN

conventional-commits

常规提交指南。创建结构化、语义化的提交记录。提交类型包括:feat(新功能)、fix(修复缺陷)、docs(文档更新)、style(代码风格调整)、refactor(重构)、perf(性能优化)、test(测试)、build(构建)、ci(持续集成)、chore(日常维护)、revert(撤销)。

SKILL.md
--- frontmatter
name: conventional-commits
# IMPORTANT: Keep description on ONE line for Claude Code compatibility
# prettier-ignore
description: "Conventional Commits guidance. Create structured, semantic commits. Types include feat, fix, docs, style, refactor, perf, test, build, ci, chore, revert."

Conventional Commits

Quick Start

Format: type(scope?): subject + body + footer

Types: feat (feature), fix (bug), docs, style, refactor, perf, test, build, ci, chore, revert

Rules: Lowercase type, imperative mood ("add" not "added"), limit subject to 50 chars

bash
feat(api): add user authentication endpoint

Implement OAuth2 flow with token refresh support.

Closes #123

Quick Reference

TypeUsageExample
featNew featurefeat(auth): add OAuth login
fixBug fixfix(api): resolve token expiration
docsDocumentationdocs(readme): update install steps
styleFormattingstyle(components): fix indentation
refactorCode changerefactor(utils): simplify parser
perfPerformanceperf(api): add response caching
testTeststest(auth): add unit tests
buildBuild systembuild(gradle): update deps
ciCI configci(github): add workflow
choreMiscchore(deps): bump lodash
revertRevertrevert: feat(api): old feature

Reference Files

Notes

  • Use imperative mood: "add feature" not "added feature" or "adds feature"
  • Scope is optional but recommended for mono-repos
  • Breaking changes: add ! after type/scope: feat(api)!: remove legacy endpoint
  • Last verified: 2025-01-17
<!-- PROGRESSIVE DISCLOSURE GUIDELINES: - Keep this file ~50 lines total (max ~150 lines) - Use 1-2 code blocks only (recommend 1) - Keep description <200 chars for Level 1 efficiency - Move detailed docs to references/ for Level 3 loading - This is Level 2 - quick reference ONLY, not a manual LLM WORKFLOW (when editing this file): 1. Write/edit SKILL.md 2. Format (if formatter available) 3. Run: claude-skills-cli validate <path> 4. If multi-line description warning: run claude-skills-cli doctor <path> 5. Validate again to confirm -->