Token导航 LogoToken导航TokenDH.com
前端设计只读github未标认证来源可访问许可证需确认审计通过

commit-creator提交创建者

Agent Skill

commit-creator 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

1,058

周安装

45

GitHub Stars

公开资料未说明

下载量

371
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:commit-creator(提交创建者)
来源仓库:https://github.com/perdolique/workflow
仓库路径:skills/commit-creator
安装命令:
npx skills add https://github.com/perdolique/workflow --skill commit-creator
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/perdolique/workflow --skill commit-creator

简介

commit-creator 生成结构化的提交消息,强制使用英文与 imperative mood 描述。

  • 适用于 monorepo 项目,要求精确指定 scope 为包名或 all ,避免歧义。
  • body 部分采用 bullet list 格式,每条对应一个逻辑变更,禁止合并多项到一行。
  • 输出不包含时间戳或作者信息,应由用户自行追加 --author 参数或配置 git config。
  • 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

Code Committing

Format

Language Requirement

Always write in English only

<type>(<scope>): summary
  • Summary: ≤50 chars, imperative mood, no period
  • Scope: Module/package name (monorepo: exact package name or all)
  • Body (optional): Bullet list - {emoji} {text}. Aim to keep explanatory bullets within about 100 characters when that preserves readability, but allow longer lines for package/version entries, URLs, or other tokens that become awkward when wrapped. One bullet = one logical change. Do not group multiple items on a single line or leave empty lines between bullets unless you intentionally want separate paragraphs.
  • Breaking: Add ! after type and BREAKING CHANGE: footer
  • Issues: End the body with a bullet like - Fixes #123 or - Fixes PROJ-456
  • No co-authorship: Never add Co-authored-by:, Co-Authored-By:, or any attribution to Copilot, coding agents, AI assistants, or any other automated tools.

Dependency update details

When the commit includes dependency or package version updates, spell out every updated package in the body.

  • Use one bullet per updated package
  • Include both old and new version for each package
  • Use the format - 📦 package-name: old-version -> new-version
  • Never collapse multiple updates into vague text like updated dependencies or bump packages
  • If several packages changed, list all of them separately

Message assembly

Git treats every extra -m flag as a separate paragraph and inserts a blank line between paragraphs.

  • If the commit has only a summary, use a single -m
  • If the commit has a body, use exactly one additional -m for the entire body, with plain newlines between bullet lines
  • Never use one -m flag per bullet, per package update, or per issue reference
  • If multiline quoting becomes awkward, write the whole message with git commit -F or git commit -F-

Keep the final message layout like this:

<type>(<scope>): summary

- first bullet
- second bullet
- third bullet

Types: feat ✨, fix 🐛, docs 📚, style 💄, refactor ♻️, perf ⚡, test ✅, build 🔧, ci 👷, chore 🔨, revert ⏪

Workflow

Branch check

Before committing, check the current branch:

git branch --show-current
git remote show origin | grep "HEAD branch"

If the current branch is the default branch (e.g. master, main) and the user has not explicitly indicated they want to commit to it, ask the user:

  • Create a new branch and commit there
  • Commit directly to the current (default) branch

Wait for user's answer before proceeding.

If the user explicitly stated the target branch in their request (e.g. "commit to master", "commit here"), skip the question and proceed.

Staging behavior

When both staged and unstaged changes exist in the working directory, and interaction is available:

  • Ask the user whether to:

- Stage all files before committing - Commit only the currently staged changes

Running git commit

After executing git commit, wait for the process to exit on its own — do not interrupt or kill it. Pre-commit hooks (linters, type checkers, test runners) can run for a long time without producing any output. Killing the process mid-run causes an exit code 130 (SIGINT) and leaves the working tree in a dirty state.

When the commit body spans multiple lines, prefer one of these command shapes:

git commit -m "chore(deps): update development dependencies" -m "- 📦 eslint: 8.57.0 -> 9.0.0
- 📦 prettier: 2.8.8 -> 3.0.0
- 🔧 Update lint configuration for new versions"
git commit -F- <<'EOF'
chore(deps): update development dependencies

- 📦 eslint: 8.57.0 -> 9.0.0
- 📦 prettier: 2.8.8 -> 3.0.0
- 🔧 Update lint configuration for new versions
EOF

Do not generate git commit -m "summary" -m "bullet 1" -m "bullet 2" unless you explicitly want blank lines between those body paragraphs.

Commit error handling

Exit code 130 (interrupted):

The commit process was interrupted — this is not a validation failure. Do not auto-retry. Report that the commit was interrupted and ask the user whether to:

  • Try again
  • Cancel

Any other non-zero exit code (validation failure):

If the commit fails (e.g., due to pre-commit hooks, linting failures, or other validation errors):

  • Report the exact error message and reasons for the failure
  • Ask the user whether to:

- Commit with --no-verify flag to bypass hooks - Attempt to fix the issues automatically - Let the user fix the issues manually

Examples

Simple feature:

feat(button): add loading state

- ✨ Add spinner icon during async operations
- 📦 @ui/icons: v1.0.0 → v1.1.0
- Fixes #42

Breaking change:

feat(theme)!: redesign color tokens

- ✨ Replace RGB values with HSL format
- 💄 Update all component styles to use new tokens
- 📦 @ui/theme: v2.1.0 → v3.0.0

BREAKING CHANGE: Color token values changed from RGB to HSL format

For more examples, see references/examples.md

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

Claude

33.57%
按下载量换算125

Codex

32.56%
按下载量换算121

Cursor

20.78%
按下载量换算77

Gemini CLI

8.62%
按下载量换算32

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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