
Log Analyzer
When you're staring at thousands of log lines trying to figure out why production is broken, this skill automates the grunt work. It detects common log formats (nginx, Spring Boot, JSON, syslog), clusters similar errors so you're not reading the same stack trace 89 times, and looks for timing patterns that suggest root causes. The workflow is solid: it filters by time and severity, extracts response times and status codes, then generates a markdown report with top errors and potential causes. The best part is the causal chain analysis, where it tries to connect a database timeout to the cascade of NPEs that followed. Just watch out for large files since it processes everything in memory.
- GitHub stars
- 328
- last updated
- May 13, 2026
Install
$ npx -y skills add laolaoshiren/claude-code-skills-zh --skill log-analyzer --agent claude-codeInstalls into .claude/skills of the current project.
Files
📋 log-analyzer — 日志分析助手
触发条件
当用户要求以下操作时激活此技能:
- 分析应用日志文件(.log, .txt, journalctl 输出等)
- 排查线上问题和异常
- 从日志中提取统计信息
- 生成日志分析报告
工作流程
1. 日志格式识别
自动检测日志格式:
- 结构化: JSON logs, Logfmt
- 常见格式: Nginx/Apache access log, Spring Boot, Django, Rails
- 系统日志: syslog, journalctl, Windows Event Log
- 自定义格式: 根据正则提取字段
2. 日志预处理
- 按时间范围过滤(如「最近 1 小时」「今天」)
- 按日志级别筛选(ERROR, WARN, INFO, DEBUG)
- 按来源/模块/服务过滤
- 去重和聚合相似错误
3. 异常模式检测
- 错误聚类: 将相似错误归为一组,避免重复分析
- 频率分析: 识别高频错误和突发异常
- 时间关联: 发现错误是否与特定事件(部署、流量高峰)相关
- 因果链追踪: 从根因错误到最终表现
4. 关键信息提取
- 错误堆栈分析(定位源码位置)
- HTTP 状态码分布(4xx/5xx 比例)
- 响应时间统计(P50/P95/P99)
- 内存/CPU 异常指标
5. 生成分析报告
## 📊 日志分析报告
### 概览
- 时间范围: 2024-01-15 10:00 ~ 12:00
- 总日志量: 12,847 条
- 错误数: 234 (1.8%)
- 警告数: 567 (4.4%)
### Top 5 异常
| 排名 | 错误类型 | 次数 | 首次出现 | 最近出现 |
|------|---------|------|---------|---------|
| 1 | NullPointerException | 89 | 10:15 | 11:42 |
| 2 | ConnectionTimeout | 67 | 10:23 | 11:58 |
### 根因分析
最可能的原因:数据库连接池耗尽(线索:连续出现 ConnectionTimeout 后跟大量 NullPointerException)
### 建议
1. 检查数据库连接池配置(当前可能 maxActive 太小)
2. 排查是否存在连接泄漏
3. 添加连接池监控告警
Sponsors
Details
- Repository
- github.com/laolaoshiren/claude-code-skills-zh
- License
- MIT
- First seen
- Jun 11, 2026