performance-audit

j4xie's avatarfrom j4xie

分析后端性能问题，检测 N+1 查询、无分页查询、全表扫描、AI 服务超时等。使用此 Skill 来评估系统性能瓶颈、检测数据库查询问题、或优化 API 响应时间。

⭐0stars🔀1forks📁View on GitHub🕐Updated Jan 10, 2026
[Code Review]

When & Why to Use This Skill

This Claude skill is a specialized backend performance auditing tool designed to identify and resolve critical system bottlenecks. It automates the detection of N+1 query patterns, unpaged database fetches, inefficient full-table scans, and blocking AI service calls. By scanning source code with precision, it provides actionable insights to optimize API response times, improve database scalability, and ensure robust system architecture.

Use Cases

  • Detecting N+1 query issues in loop structures to reduce redundant database round-trips and improve latency.
  • Identifying missing pagination in repository 'findAll' calls to prevent OutOfMemory (OOM) errors and high memory consumption.
  • Locating inefficient SQL 'LIKE' patterns that bypass database indexes and cause performance-degrading full table scans.
  • Auditing AI and third-party service integrations to identify synchronous blocking calls and missing timeout configurations.
  • Performing automated performance health checks on legacy Java/Spring Boot codebases to generate optimization reports.
nameperformance-audit
description分析后端性能问题，检测 N+1 查询、无分页查询、全表扫描、AI 服务超时等。使用此 Skill 来评估系统性能瓶颈、检测数据库查询问题、或优化 API 响应时间。

后端性能审计 Skill

性能问题分类

问题类型 严重程度 影响
N+1 查询 高 100 用户 = 101 次 DB 查询
无分页 findAll() 高 全表加载，内存溢出
LIKE %keyword% 中 无法使用索引
AI 服务同步阻塞 高 线程池耗尽

已知问题位置

N+1 查询

文件 行号 问题
MobileServiceImpl.java 990-995 人员统计循环查询
MobileServiceImpl.java 1061-1064 工时排行循环查询
MobileServiceImpl.java 1141-1142 加班统计循环查询
MobileServiceImpl.java 1228-1231 人员绩效循环查询

无分页 findAll()

文件 行号
PlatformServiceImpl 50-52
AIReportScheduler 61-63
FactoryServiceImpl 38-40

LIKE 全表扫描

文件 行号
CustomerRepository 49
SupplierRepository 49
MaterialBatchRepository 126
UserRepository 73-75

检测命令

cd /Users/jietaoxie/my-prototype-logistics/backend-java

# 1. N+1 查询 - for 循环中的 repository 调用
grep -rn "for.*{" src/main/java/com/cretas/aims/service/ -A10 | \
  grep -B5 "repository\|Repository" | head -50

# 2. 无分页 findAll()
grep -rn "\.findAll()" src/main/java/ --include="*.java" | grep -v "Pageable"

# 3. 左模糊 LIKE (无法使用索引)
grep -rn "LIKE %:" src/main/java/ --include="*.java"

# 4. RestTemplate 超时配置
grep -rn "RestTemplate" src/main/java/ --include="*.java" -A5 | head -30

# 5. 巨型类 (>500行)
wc -l src/main/java/com/cretas/aims/service/impl/*.java | awk '$1 > 500'

# 6. 缓存/异步使用
grep -rn "@Cacheable\|@Async" src/main/java/ --include="*.java"

修复要点

问题 修复方案
N+1 查询 批量查询 findByUserIdIn(userIds) + groupingBy
无分页 findAll(PageRequest.of(0, 100))
左模糊 改为右模糊 LIKE CONCAT(:keyword, '%')
AI 超时 setConnectTimeout(5000) + @Async

参考

  • 完整审计报告: backend-java/BACKEND_AUDIT_REPORT.md