Skip to main content

工作流快速开始

本教程将创建一个自动化 GitHub Pull Request 审查的工作流，展示如何结合 CLI 工具、文件分析和用户交互。

前置条件​

  • 已安装 Careti
  • 已安装并认证 GitHub CLI(gh)
  • 打开一个包含待测试 PR 的 Git 仓库

创建 PR 审查工作流​

该工作流会获取 PR 信息、分析改动并生成审查评论。

创建工作流文件

创建项目级工作流目录。

  1. 在项目根目录创建 .agents/context/workflows
  2. 在该目录下创建 pr-review.md

编写工作流内容

将以下内容写入 pr-review.md。

# Pull Request Reviewer

This workflow helps me review a pull request by analyzing the changes and drafting a review.

## 1. Gather PR Information
First, I need to understand what this PR is about. I'll fetch the title, description, and list of changed files.

```bash
gh pr view PR_NUMBER --json title,body,files
```

## 2. Examine Modified Files
Now I will examine the diff to understand the specific code changes.

```bash
gh pr diff PR_NUMBER
```

## 3. Analyze Changes
I will analyze the code changes for:
*   **Bugs:** Logic errors or edge cases.
*   **Performance:** Inefficient loops or operations.
*   **Security:** Vulnerabilities or unsafe practices.

## 4. Confirm Assessment
Based on my analysis, I will present my findings and ask how you want to proceed.

```xml
<ask_followup_question>
  <question>I've reviewed PR #PR_NUMBER. Here is my assessment:

[Insert Analysis Here]

Do you want me to approve this PR, request changes, or just leave a comment?</question>
  <options>["Approve", "Request Changes", "Comment", "Do nothing"]</options>
</ask_followup_question>
```

## 5. Execute Review
Finally, I will execute the review command based on your decision.

```bash
# If approving:
gh pr review PR_NUMBER --approve --body "Looks good to me! [Summary of analysis]"

# If requesting changes:
gh pr review PR_NUMBER --request-changes --body "Please address the following: [Issues list]"

# If commenting:
gh pr review PR_NUMBER --comment --body "[Comments]"
```
ℹ️Note

运行时请将 PR_NUMBER 替换为实际 PR 编号（例如：/pr-review.md 123）。

运行工作流

现在可以运行该工作流了。

  1. 打开 Careti 聊天面板
  2. 输入 /pr-review.md + PR 编号（例如：/pr-review.md 42）
  3. Careti 会获取 PR 信息、分析代码并展示结果
💡Tip

当 Careti 执行 gh pr view 等命令时可能会暂停显示输出，请点击 Proceed While Running 继续。

其他常见用法​

这只是一个示例。你还可以为很多任务创建工作流：

  • 创建组件: 自动生成 React 组件或 API 端点的模板
  • 运行测试: 执行测试并汇总结果
  • 自动部署: 使用 docker、kubectl 等工具
  • 分步重构: 引导完成复杂重构

在你的项目中寻找重复性工作，并将它们转化为高效的 Careti 工作流。