Token导航 LogoToken导航

Agent Skill · 开发

dev%3acommit(dev%3a 提交)

dev%3acommit 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。

dev%3acommit 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,暂无第三方扫描结果,按安装量在开发类的 21,746 个 Skill 里排第 19,191。

总安装 288周安装 12GitHub Stars 公开资料未说明下载 96核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:dev%3acommit(dev%3a 提交)
来源仓库:https://github.com/cloudvoyant/codevoyant
仓库路径:skills/dev%3Acommit
安装命令:
npx skills add https://github.com/cloudvoyant/codevoyant --skill dev:commit
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/cloudvoyant/codevoyant --skill dev:commit

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

用于处理 GitHub 仓库、Issue 和 Pull Request 协作信息,适合跟踪代码变更。可结合来源仓库和原始 README 核验具体用法和协作流程。安装前建议确认权限范围和维护状态,避免触发不必要的网络请求。适用于需要围绕仓库状态进行整理的场景,如代码审核或协作管理。

dev%3acommit 属于开发类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/dev%3Acommit

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/cloudvoyant/codevoyant --skill dev:commit;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

dev%3acommit 怎么安装?

在终端执行 npx skills add https://github.com/cloudvoyant/codevoyant --skill dev:commit。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

dev%3acommit 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

dev%3acommit 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

dev%3acommit 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

dev%3acommit 用的人多吗?

公开数据是累计安装 288 次,近一周安装 12 次,下载 96 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

dev%3acommit 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 要求有安全扫描结论才肯安装的团队。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
dev%3acommit

dev%3a 提交 · 本页

288—接触敏感数据未明确未收录

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Compatibility: If AskUserQuestion is unavailable, present options as a numbered list and wait for the user's reply. If Task is unavailable, run parallel steps sequentially. The context: fork and agent: frontmatter fields are Claude Code-specific — on OpenCode and VS Code Copilot they are ignored and the skill runs inline using the current model.

Create a git commit following conventional commit standards with a professional, concise message.

Rules

Choosing the right type

  1. Does it add new functionality users can use? → feat
  2. Does it fix broken behavior? → fix
  3. Does it change existing behavior (breaking)? → feat! or fix!
  4. Does it only improve code structure? → refactor
  5. Does it only update documentation? → docs
  6. Does it only affect tests? → test
  7. Does it only affect build/CI? → chore
Bump behavior is determined by your project's release config (.releaserc.*, commitizen.config.*, [tool.commitizen]). Do not predict version bumps — check those files if needed.

Never commit secrets

  • Secrets or credentials (.env files, API keys, passwords)
  • Warn user if such files are staged

Formatting

  • Always use HEREDOC format for commit messages to ensure proper formatting.
  • First line max 72 characters
  • Use imperative mood: "add feature" not "added feature" or "adds feature"
  • No period at end of first line
  • Be professional and concise
  • Do NOT include self-attribution (no "Generated with Claude Code", no "Co-Authored-By: Claude")
  • Body: use bullet points, not prose paragraphs. Each bullet is one change or reason. Keep bullets terse — one line each, under 72 chars.
  • Body is optional — only include if the changes aren't obvious from the subject line
  • Make body as short as possible, as few bullets as possible
  • Use type(scope) format when changes touch a distinct subsystem — especially with --atomic commits. Scope groups entries in the changelog under that feature/fix category (e.g. feat(opencode), fix(vscode)).

Scoped vs. unscoped

  • Use feat(scope) / fix(scope) when the change is clearly bounded to one area: a plugin, tool, subsystem, or named feature
  • Use plain feat: / fix: for cross-cutting changes with no single scope
  • Scope should be short, lowercase, no spaces: opencod, vscode, spec, help, e2e, docs, install

Flags

  • --yes or -y: Skip commit message confirmation (auto-approve message)
  • --no-push: Commit only — do not push or monitor CI
  • --autofix: After push, if CI fails, automatically attempt to fix failures and re-push
  • --atomic: Detect logical change groups and create one commit per group
  • --single (default): All staged changes in one commit

Workflow

Step 0: Parse Flags

YES=false; NO_PUSH=false; AUTOFIX=false; ATOMIC=false
[[ "$*" =~ --yes|-y ]]    && YES=true
[[ "$*" =~ --no-push ]]   && NO_PUSH=true
[[ "$*" =~ --autofix ]]   && AUTOFIX=true
[[ "$*" =~ --atomic ]]    && ATOMIC=true

Step 1: Check Git Status (Fast Path)

Use conversation context first - the changes are usually already in context from the work session.

Only run git commands if you need to verify:

git status
git diff --stat

Skip git log - you should already know the commit message style from:

  • Previous commits in this session
  • Project's CLAUDE.md conventions
  • Standard conventional commit format

Skip reading plan.md - if you just implemented a plan, it's already in context.

Only run additional commands if you're truly uncertain about what changed.

Step 1.5: Format and Lint

Run formatters and linters before staging so any auto-fixes are included in the commit.

Formatters (auto-fix, run unconditionally if available):

npx @codevoyant/agent-kit task-runner run format 2>/dev/null || true

If formatter ran and modified files: report ✓ Formatter applied — changes will be included in commit.

Linters (report errors, block commit if they fail):

npx @codevoyant/agent-kit task-runner run lint 2>/dev/null || \
npx @codevoyant/agent-kit task-runner run check 2>/dev/null || true

If linting fails: report the errors and stop — do not proceed to staging until fixed:

✗ Linting failed — fix the errors above before committing.

Skip silently if no formatter or linter is configured.

Step 2: Review with User

If ATOMIC is true:

Analyze staged file paths and detect logical groups (by component, package, or directory). Draft a commit message for each group. Output all proposed commits:

Proposed commits:

1. feat(auth): add JWT token validation
2. docs: update API reference
3. test(auth): add token expiry tests

Then ask for confirmation using the AskUserQuestion tool:

questions:
  - question: 'Do these commits look good?'
    header: 'Review Commits'
    multiSelect: false
    options:
      - label: 'Looks good — commit all'
        description: '{N} commits'
      - label: 'Cancel'
        description: "Don't commit"
  • If "Looks good — commit all": proceed to Step 3.
  • If "Cancel": exit without committing.
  • If Other (user edits): apply their changes and proceed to Step 3.

If SINGLE (default):

Output the full proposed commit message:

Proposed commit message:

{full commit message, including body if present}

Then ask for confirmation using the AskUserQuestion tool:

questions:
  - question: 'Does this commit message look good?'
    header: 'Review Commit'
    multiSelect: false
    options:
      - label: 'Looks good — commit'
        description: '{first line of proposed message}'
      - label: 'Cancel'
        description: "Don't commit"
  • If "Looks good — commit": proceed to Step 3.
  • If "Cancel": exit without committing.
  • If Other (user typed a custom message): use it as-is and proceed to Step 3.

If AUTO_APPROVE is true: skip this prompt entirely and proceed directly to Step 3. Report: ✓ Auto-approved with --yes flag.

Step 3: Stage and Commit

If ATOMIC is true: create one commit per group in order:

git add <files-in-group> && git commit -m "$(cat <<'EOF'
<type>(<scope>): <description for this group>
EOF
)"

Default (SINGLE):

git add -A && git commit -m "$(cat <<'EOF'
<type>: <description>

[optional body]
EOF
)"

Step 5: Push and Verify CI

If NO_PUSH is true: skip this step entirely and report ✓ Committed (push skipped).

Otherwise — always push, then launch CI monitoring in background:

  1. Push: git push origin <branch>
  2. After push succeeds, launch CI monitoring as a background Task — do NOT wait for it:
Agent:
  subagent_type: general-purpose
  run_in_background: true
  description: "CI monitoring"
  prompt: "Run /dev:ci [--autofix if AUTOFIX=true]. Monitor CI and report results when done."
  1. Report immediately: ✓ Committed and pushed. CI monitoring running in background — you'll be notified when checks complete.

Skip CI monitoring if:

  • Repo has no CI workflows configured
  • Neither gh nor glab CLI is installed (inform but don't block)

平台分布

Claude

32.95%
按下载量换算32

Codex

32.65%
按下载量换算31

Cursor

19.77%
按下载量换算19

Gemini CLI

10.28%
按下载量换算10

安全审计

暂无安全审计结果可展示。

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills