pr-address-bot-reviews

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

PR Address Bot Reviews

处理PR机器人评审反馈

Triage and resolve bot-generated PR feedback with a consistent workflow.

通过一致的工作流分类并解决机器人生成的PR反馈。

Supported Sources

支持的来源

```
coderabbitai[bot]
```
```
chatgpt-codex-operator
```

Add more bot logins when needed.

```
coderabbitai[bot]
```
```
chatgpt-codex-operator
```

需要时可添加更多机器人登录账号。

Workflow

工作流

1. Collect Feedback

1. 收集反馈

Identify the target PR number (from current branch PR or user input).
Fetch PR review data with
```
gh
```
:
- reviews
- review comments
- issue comments
Filter feedback to supported bot authors.

确定目标PR编号（从当前分支PR或用户输入获取）。
使用
```
gh
```
获取PR评审数据：
- 评审记录
- 评审评论
- 议题评论
筛选出支持的机器人作者的反馈。

2. Normalize and Group

2. 标准化与分组

Deduplicate repeated suggestions.
Group by file/path and concern type:
- correctness/bug risk
- security
- performance
- maintainability/style
- test/documentation gaps

去重重复的建议。
按文件/路径和问题类型分组：
- 正确性/ bug风险
- 安全性
- 性能
- 可维护性/风格
- 测试/文档缺失

3. Decide Actionability

3. 判定可执行性

Classify each item into:

```
must-fix
```
: clear bug/risk/regression
```
should-fix
```
: strong improvement with low downside
```
skip-with-reason
```
: incorrect assumption, out-of-scope, or high-cost low-value
```
needs-user-decision
```
: product/architecture tradeoff

将每个项分类为：

```
必须修复
```
：明确的bug/风险/回归问题
```
建议修复
```
：提升效果显著且负面影响小
```
跳过并说明原因
```
：假设错误、超出范围或高成本低价值
```
需用户决策
```
：产品/架构权衡问题

4. Implement Needed Changes

4. 实施必要变更

Apply
```
must-fix
```
and agreed
```
should-fix
```
items.
Keep changes scoped to the PR intent.
Run relevant checks/tests for touched areas.

应用
```
必须修复
```
和已确认的
```
建议修复
```
项。
保持变更范围与PR的初衷一致。
对涉及的区域运行相关检查/测试。

5. Respond on PR

5. 在PR上回复

For fixed items: reply with what changed.
For skipped items: reply with concise rationale.
For tradeoffs: ask explicit decision questions.

Reply in the correct channel:

Inline review thread: prefer

skills/pr-address-bot-reviews/scripts/reply-to-review-comment <pr> --login <bot-login> --path <file> [--line <line>] "<response>"

Lower-level flow remains available:
```
find-review-comment-id
```
then
```
pr-thread-reply
```
General PR note: use
```
gh pr comment
```

已修复项：回复说明变更内容。
跳过项：回复简洁的理由。
权衡项：提出明确的决策问题。

在正确的渠道回复：

内联评审线程：优先使用

skills/pr-address-bot-reviews/scripts/reply-to-review-comment <pr> --login <bot-login> --path <file> [--line <line>] "<response>"

仍可使用底层流程：
```
find-review-comment-id
```
然后
```
pr-thread-reply
```
通用PR备注：使用
```
gh pr comment
```

6. Report Back to User

6. 向用户反馈

Summarize by category:
- fixed
- skipped (with reason)
- pending decision
Provide changed files and verification commands run.

按类别汇总：
- 已修复
- 已跳过（含原因）
- 待决策
提供变更的文件和运行的验证命令。

Decision Heuristics

决策准则

Prefer concrete evidence over bot confidence.
Reject suggestions that conflict with project conventions.
Reject speculative refactors that increase scope without reducing risk.
Escalate when the fix changes behavior, API contract, or data model.

优先依据具体证据而非机器人的置信度。
拒绝与项目惯例冲突的建议。
拒绝增加范围却未降低风险的推测性重构。
当修复会改变行为、API契约或数据模型时，需升级处理。

Minimal Command Sequence

最简命令序列

bash

gh pr view <number> --json number,title,body,reviews,comments,files
gh pr view <number> --comments

bash

gh pr view <number> --json number,title,body,reviews,comments,files
gh pr view <number> --comments

preferred single-command inline review reply flow

lower-level two-step flow when you want explicit control

需明确控制时使用的两步底层流程

skills/pr-address-bot-reviews/scripts/find-review-comment-id <number> --login <bot-login> --path <file> [--line <line>] skills/pr-address-bot-reviews/scripts/pr-thread-reply <number> <comment_id> "<response>"

optional general PR note

可选的通用PR备注

gh pr comment <number> --body "<response>"

implement code changes

实施代码变更

git add <files...> git commit -m "<message>" git push

undefined