Review Process
1. Plan Review
Object：spec.md + tasks.md
Checklist：
- Are the requirement goals clear and unambiguous?
- Does the functional description fully cover the original requirements?
- Is the impact scope accurate (modules, interfaces, data tables, configuration)?
- Are the acceptance criteria quantifiably verifiable?
- Is the task breakdown reasonable (moderate granularity, no omissions)?
- Are the dependencies among tasks correct?
- Are there any implicit technical debts or risks?
Loop Limit: Maximum of 3 rounds; after exceeding this, escalate to manual decision-making.
Output：request_analysis/review/review_v1.md
2. Execution Review
Object: Code changes (diff relative to the baseline branch)
Checklist：
- Does it satisfy all functional requirements in spec.md?
- Does it follow the layered architecture specification?
- Are exception boundaries covered (None, out of bounds, invalid parameters)?
- Are failure/degradation scenarios handled?
- Are the price field type and precision correct?
- Are timeouts and degradation set for external calls?
- Are the logs complete (key paths, exception traceback)?
- Do internationalization/multiple environments require synchronized modifications?
- Are there any security risks (SQL injection, privilege escalation, etc.)?
Loop Limit: Maximum of 2 rounds; after exceeding this, escalate to manual decision-making.
Output：coding/review/code_review_v1.md
Review Comment Classification
[MUST FIX] 问题描述
原因：为什么必须修复
建议：如何修复
位置：文件:行号
[LOW] 问题描述
建议：可选优化方式
[INFO] 问题描述
仅供知悉，无需修改
- MUST FIX — Failure to fix will cause online failures, data errors, or severe bugs
- LOW — Recommended optimization, does not affect functional correctness
- INFO — For information only
Review Report Template
# {类型}评审报告 v{版本号}
## 概述
- 评审对象：{spec.md / 代码变更}
- 评审日期：{YYYY-MM-DD}
- 评审结论：{APPROVED / REVISION REQUIRED}
- MUST FIX：{数量} | LOW：{数量} | INFO：{数量}
- 评审轮次：{N}
## 评审意见
### MUST FIX
...
### LOW
...
### INFO
...