gitlab-mr-review-flow
标准化“需求描述 → 规范检索 (QMD) → 实现改动 → GitLab MCP 创建 MR → 代码评审报告”的流程技能。
Best use case
gitlab-mr-review-flow is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
标准化“需求描述 → 规范检索 (QMD) → 实现改动 → GitLab MCP 创建 MR → 代码评审报告”的流程技能。
Teams using gitlab-mr-review-flow should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/gitlab-mr-review-flow/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How gitlab-mr-review-flow Compares
| Feature / Agent | gitlab-mr-review-flow | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
标准化“需求描述 → 规范检索 (QMD) → 实现改动 → GitLab MCP 创建 MR → 代码评审报告”的流程技能。
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# GitLab MR Review Flow with QMD & Rules
## Overview
规范化执行从需求拆解到 MR 创建与代码评审的全流程。
**核心增强**：
1. **QMD 集成**：在开发前自动检索相关的工程规范与文档。
2. **GitLab MCP**：强制使用 MCP 工具管理 MR。
3. **中文评审报告**：按规范输出标准化报告。
## Workflow Decision Tree
- 需要检索内部规范/文档 → 使用 `qmd` skill (search/get)。
- 需要隔离环境/并行修改 → 使用 `using-git-worktrees` skill。
- 涉及任何 git 操作 → 必须使用 `git-master` skill。
- 需要创建/更新 MR → 必须使用 GitLab MCP。
- 需要评审报告 → 使用 `requesting-code-review` skill 并输出中文报告。
## Step 1. Gather Inputs & Context
收集并确认：
- 仓库路径与目标模块范围
- 需求来源（文档/issue/日志）与验收标准
- **关键词提取**：用于 QMD 检索的关键词 (e.g., "API 规范", "数据库设计", "错误处理").
## Step 2. Retrieve Norms & Documentation (QMD)
在动手写代码前，先检索相关的业务或技术规范。
1. **Search**: 根据提取的关键词检索文档。
```bash
# 示例
qmd search "API 规范"
qmd search "数据库命名"
```
2. **Read**: 读取最相关的 1-2 份文档，提取对本次开发有约束力的条款。
```bash
qmd get "path/to/relevant-doc.md"
```
3. **Context Injection**: 将提取到的规范点记录在上下文中，后续用于 Self-Check 和 Code Review。
## Step 3. Prepare Workspace & Implement
1. 读取仓库内 `AGENTS.md` 约束。
2. 如需隔离，使用 `using-git-worktrees`。
3. 实现改动，保持原子提交。
4. 执行测试并记录结果。
## Step 4. Branching & Push
### 分支命名规则
* **Target**: `fix-{feature}-mr` 或 `feat-{feature}-mr` (空分支)
* **Source**: `fix-{feature}` 或 `feat-{feature}` (含改动)
### 操作流程
```bash
# 自动创建 Target 和 Source 分支
./scripts/start_feature.sh "{feature-name}"
```
## Step 5. Create MR via GitLab MCP
必须使用 GitLab MCP。禁止浏览器操作。
**MR 描述模板** (需包含根因、改动、**以及遵循的规范**):
(Script `create_mr.sh` contains the template)
Automatically create MR using the script:
```bash
# Usage: ./scripts/create_mr.sh "MR Title"
export GITLAB_PERSONAL_ACCESS_TOKEN="<your-token>"
./scripts/create_mr.sh "feat: <Title>"
```
## Step 6. Code Review Report (中文)
调用 `requesting-code-review` skill，结合 Step 2 获取的规范进行评审。
**评审重点**:
1. 是否符合 QMD 检索到的规范？
2. 是否存在逻辑漏洞？
3. 代码风格与最佳实践。
**报告模板**:
```markdown
**AI Review:** {YYYYMMDD}-{BUILD_ID}
**参考规范库**: {列出参考的 QMD 文档}
对比基准： {BASE_BRANCH}
一、规范符合度检查 (Based on QMD)
- [ ] 规范A: {检查结果}
- [ ] 规范B: {检查结果}
二、严重问题
...
三、中等问题
...
四、总结
...
```
## Step 7. Inline Code Comments (Discussions)
对于规范违例，**必须**在 MR Diff 中进行行级评论。
1. **Locate Violation**: Identify file and line number.
2. **Post Discussion**: Run the helper script.
```bash
# Usage: ./scripts/post_comment.sh <file> <line> "<comment>"
./scripts/post_comment.sh "bad_code.py" 10 "❌ Violation: ..."
```Related Skills
jj-workflow
Jujutsu atomic workflow with full operational reference for jj-based repositories.
jj-vcs-workflow
Jujutsu (jj) VCS の総合ガイド。基本コマンド・Git移行・並列開発・履歴操作・PRレビュー・安全な push ワークフローをカバー。以下の場合に使用: (1) jj コマンドの使い方を確認したいとき (2) Git から jj への移行時 (3) 並列開発・履歴書き換え・コンフリクト解消を行うとき (4) PR レビュー対応時 (5) push を実行したいとき
@gw-git-worktree-workflows
Master Git worktrees and gw-tools workflows for parallel development. Use this skill when creating worktrees, managing multiple branches simultaneously, navigating between worktrees, troubleshooting worktree issues, or setting up feature branch workflows. Triggers on tasks involving git worktree commands, branch isolation, parallel development, or gw CLI usage.
gitlab-ci-generator
Comprehensive toolkit for generating best practice GitLab CI/CD pipelines and configurations following current standards and conventions. Use this skill when creating new GitLab CI/CD resources, implementing CI/CD pipelines, or building GitLab pipelines from scratch.
github-workflows
Initialize or update GitHub Actions workflows for Go projects with comprehensive CI/CD pipelines including linting, testing, coverage, snapshot builds, and releases. Use when setting up GitHub Actions automation for Go projects. Trigger with "setup github actions", "add github workflows", or "configure ci/cd".
github-pr-review-comments
Comprehensive workflow for managing GitHub PR review comments using gh CLI and GraphQL API. Use when asked to address review comments, find unreplied comments, reply to review threads, or resolve/unresolve review conversations. Supports finding ALL comments across pagination boundaries, replying to threads, and resolving conversations.
Gitee Workflow Automation
深度集成 Gitee MCP，实现 Issue 管理、PR 自动化提交、代码审查和版本发布的全流程自动化。
git-workflow
Git workflow guidance for commits, branches, and pull requests
git-workflow-automation
Comprehensive Git workflow automation including branching strategies, pull request creation, code reviews, merge strategies, and release management. Use when Claude needs to help with Git operations, branching models (Git Flow, GitHub Flow), pull request creation, code reviews, merge conflicts, or release processes.
flow-interview
Interview user in-depth about a spec, bead, or feature idea to extract complete implementation details. Use when user wants to flesh out a spec, refine requirements, discuss technical details, clarify a feature before building, or asks to be interviewed about their idea. Triggers on "interview me", "ask me questions about", "flesh out this spec", "refine requirements", "help me think through".
differential-review
Perform security-focused review of code diffs and pull requests, identifying newly introduced vulnerabilities, security regressions, and unsafe patterns in changed code.
codex-code-review
Perform comprehensive code reviews using OpenAI Codex CLI. This skill should be used when users request code reviews, want to analyze diffs/PRs, need security audits, performance analysis, or want automated code quality feedback. Supports reviewing staged changes, specific files, entire directories, or git diffs.