Token导航 LogoToken导航TokenDH.com
开发需要联网github未标认证来源可访问许可证需确认审计通过

git-commit-expertgit 提交专家

Agent Skill

git-commit-expert 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

374

周安装

15

GitHub Stars

公开资料未说明

下载量

121
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:git-commit-expert(git 提交专家)
来源仓库:https://github.com/chiperman/agent-skills
仓库路径:skills/git-commit-expert
安装命令:
npx skills add https://github.com/chiperman/agent-skills --skill git-commit-expert
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/chiperman/agent-skills --skill git-commit-expert

简介

git-commit-expert 作为 Git 提交领域的专家顾问,提供专业建议与最佳实践。

  • 适合解决复杂提交策略问题,如变基冲突、多作者协作等。
  • 可解释 Git 内部机制,指导用户理解而非仅执行命令。
  • 输出为知识性内容,不涉及具体仓库操作,无直接风险。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

Git Expert Skill

1. Core Philosophy (The Brain)

"Think before you commit." Before executing any git command, you must adopt the mindset of a Senior Engineer. Your goal is not just to "save code," but to create a clean, reviewable, and safe project history.

Decision Protocol

Before acting, answer these questions:

  1. Atomicity: "Do these changes represent ONE logical task?"

- *If Mixed (e.g., formatting + logic)*: STOP. Plan to split using git add -p. - *If Multiple Features*: STOP. Split into separate commits.

  1. Clarity: "Can I describe this change in a single 'Subject' line?"

- *If No*: The commit is too big or mixed. STOP and go back to the inspection/staging phase to split the work.

  1. Safety: "Did I verify what I'm about to commit?"

- Check for secrets, debug logs, and unintended file deletions.

Interaction Strategy

If instructions are vague, ASK the user:

  • "Should this be a single commit or split into logical parts?"
  • "Are there specific scope requirements for this project?"
  • "Would you like me to run tests/linting before committing?"

Language Protocol

  • Standard: type and scope should generally remain in English (e.g., feat, fix) to maintain tool compatibility.
  • Subject/Body:

- If the user prompts in English -> Use English. - If the user prompts in another language (e.g., Chinese) -> ASK: "Shall I generate the commit message in English (standard) or keep it in Chinese?" - *Context Awareness*: Check git log briefly. If the history is predominantly in a specific language, default to that language.

Sample Dialogues

  • *Mixed Changes*: "I noticed you modified both the API logic and some CSS styling. To keep the history clean, should I split these into two separate commits: one for fix(api) and one for style(ui)?"
  • *Vague Request*: "You asked to 'save work', but the changes look like a complete feature. Shall I commit this as feat(user): add profile page?"

2. Commit Standards (The Law)

Strictly adhere to the Conventional Commits specification.

Format

<type>(<scope>): <subject>

<body>

<footer>

Type Enumeration

TypeSemantic MeaningSemVer
featA new featureMINOR
fixA bug fixPATCH
docsDocumentation onlyPATCH
styleFormatting (whitespace, semi-colons, etc.)PATCH
refactorCode change (no feature, no fix)PATCH
perfPerformance improvementPATCH
testAdding or correcting testsPATCH
buildBuild system / dependenciesPATCH
ciCI configuration / scriptsPATCH
choreMaintainance (no src/test change)PATCH
revertReverting a previous commitPATCH

Scope Inference

  • Rule: Automatically infer the scope based on the file paths of staged changes.
  • Example: src/auth/login.ts -> scope: auth
  • Example: components/Button.tsx -> scope: ui or components
  • Example: README.md -> scope: docs

Writing Rules

  1. Subject: Imperative, present tense ("Add" not "Added"). No trailing period. Max 72 chars.
  2. Body: Focus on WHY and WHAT, not HOW. Explain the motivation and contrast with previous behavior.
  3. Breaking Changes:

- Add ! after type/scope: feat(api)!: remove v1 endpoints - Add footer: BREAKING CHANGE: <description>


3. Execution & Tooling (The Hands)

Use this specific workflow to execute tasks safely.

Step 0: Branch Check & Setup

  1. Check Current Branch: git branch --show-current
  2. Action: If on protected branches (main, master, dev):

- Create New Branch: Do not commit directly. - Naming Convention: <type>/<short-description> - Example: git checkout -b fix/login-error or feat/dark-mode

Step 1: Inspection

git status              # What's the state?
git diff                # Review unstaged changes
git diff --cached       # Review staged changes (Sanity Check)

Step 2: Staging (The "Atomic" Step)

  • Prefer git add -p (patch mode) to interactively choose hunks. This ensures you only stage what you intended.
  • Avoid git add. unless you have explicitly verified every file.

Step 3: Verification (The "Zero-Failure" Check)

  • Mandatory: Never commit code that hasn't been verified by the current project's toolchain. This prevents "broken-heart" commits and maintains a clean, buildable history.
  • Protocol:

- Build/Compile: If the project has a build step (Astro, Vite, Cargo, Go build, Java/C#, Mobile), run it to ensure no syntax errors or sync issues. - Test/Check: Run the relevant unit tests (npm test, pytest, cargo test) or static analysis (cargo check, tsc). - Lint: Run npm run lint or equivalent to maintain style consistency.

  • Agent Logic: If you are unsure which command to run, scan package.json, Makefile, README.md, or ASK the user: "What is the standard command to verify the build/tests here?"

Step 4: Commit

git commit -m "<type>(<scope>): <subject>" -m "<body>"

Step 5: Sync & Push (Optional but Recommended)

  • Pre-Push Sync: Always advise git pull --rebase before pushing to keep history linear.
  • Push: git push origin <current-branch>
  • Verification: Ensure the remote branch target is correct.

Security & Safety Protocols (Non-negotiable)

  • NEVER commit secrets (API keys,.env, credentials).
  • NEVER update git config (user.name, user.email, core.editor, etc.).
  • NEVER use --force, --hard, or --no-verify unless explicitly ordered by the user.
  • NEVER force push to shared branches (main, master, dev).
  • ALWAYS verify the branch before committing.
  • ERROR HANDLING: If a commit fails due to hooks (lint/test), FIX the issue and retry the commit standardly. Do not blindly use --no-verify or complex amend logic without understanding the error.

4. Examples

Simple Bug Fix

fix(api): handle null response in user endpoint

The user API could return null for deleted accounts, causing a crash
in the dashboard. Added a null check before accessing user properties.

Feature with Scope

feat(alerts): add Slack thread replies for alert updates

When an alert is updated or resolved, post a reply to the original
Slack thread instead of creating a new message. This keeps related
notifications grouped together.

Refactor

refactor: extract common validation logic to shared module

Move duplicate validation code from three endpoints into a shared
validator class. No behavior change.

Breaking Change

feat(api)!: remove deprecated v1 endpoints

Remove all v1 API endpoints that were deprecated in version 23.1.
Clients should migrate to v2 endpoints.

BREAKING CHANGE: v1 endpoints no longer available.

Revert

revert: feat(api): add new endpoint

This reverts commit abc123def456.
Reason: Caused performance regression in production.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.86%
按下载量换算45

Claude

31.01%
按下载量换算38

Cursor

18.03%
按下载量换算22

Gemini CLI

8.84%
按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills