claude-reflect

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Claude Reflect Skill

Automated daily reflection analyzing git history across all workspace-hub submodules, extracting patterns via RAGS loop, and auto-creating skills from recurring patterns.

每日自动化复盘，分析workspace-hub所有子模块的Git历史，通过RAGS循环提取模式，并从重复模式中自动创建Skill。

Quick Start

快速开始

bash

undefined

bash

undefined

Run full RAGS loop manually

手动运行完整RAGS循环

/reflect

Quick 7-day reflection

快速7天复盘

/reflect --days 7

Preview patterns without creating skills

预览模式但不创建Skill

DRY_RUN=true /reflect

Force weekly report generation

强制生成周报

WEEKLY_REPORT=true /reflect

undefined

WEEKLY_REPORT=true /reflect

undefined

Automated Execution

自动执行

Cron Schedule: Daily at 5:00 AM

Runs full RAGS loop automatically
Generates weekly reports on Sundays
Auto-creates skills when pattern score >= 0.8

Cron调度： 每日凌晨5点

自动运行完整RAGS循环
周日自动生成周报
当模式评分≥0.8时自动创建Skill

When to Use

使用场景

Automated: Runs daily via cron - no manual intervention needed
Manual: Run
```
/reflect
```
to trigger immediate analysis
Before planning new features to identify reusable patterns
After major releases to capture learnings

自动执行：通过Cron每日运行，无需人工干预
手动触发：运行
```
/reflect
```
立即启动分析
规划新功能前，识别可复用模式
重大版本发布后，总结经验

Distinction from Similar Skills

与同类Skill的区别

Skill	Trigger	Scope	Data Source
`skill-learner`	Post-commit	Single repo	Last commit
`claude-reflection`	Auto/session	User interactions	Conversation
`claude-reflect`	Manual/scheduled	All 26 repos	30-day git history

Skill	触发方式	范围	数据源
`skill-learner`	提交后触发	单一仓库	最后一次提交
`claude-reflection`	自动/会话触发	用户交互	对话内容
`claude-reflect`	手动/定时触发	全部26个仓库	30天Git历史

Prerequisites

前置条件

Git access to all workspace-hub submodules
```
~/.claude/state/
```
directory for state persistence
```
~/.claude/memory/
```
directory for global patterns

拥有所有workspace-hub子模块的Git访问权限
用于状态持久化的
```
~/.claude/state/
```
目录
用于存储全局模式的
```
~/.claude/memory/
```
目录

Overview

概述

This skill analyzes git history across all 26+ workspace-hub submodules to extract development patterns and automatically enhance or create skills based on findings.

该Skill会分析所有26+个workspace-hub子模块的Git历史，提取开发模式，并基于分析结果自动优化或创建Skill。

Quick Reference

快速参考

Commands

命令

Command	Description
`/reflect`	Run reflection with default 30-day window
`/reflect --days 7`	Quick 7-day reflection
`/reflect --days 90`	Extended quarterly reflection
`/reflect --repo <name>`	Single repository reflection
`/reflect --dry-run`	Preview patterns without creating skills

命令	描述
`/reflect`	使用默认30天窗口运行复盘
`/reflect --days 7`	快速7天复盘
`/reflect --days 90`	扩展季度复盘
`/reflect --repo <name>`	单一仓库复盘
`/reflect --dry-run`	预览模式但不创建Skill

Core Workflow: RAGS Loop

核心工作流：RAGS循环

1. REFLECT - Collect Git History

1. REFLECT - 收集Git历史

Enumerate and analyze git activity across all submodules:

bash

undefined

枚举并分析所有子模块的Git活动：

bash

undefined

Enumerate submodules

枚举子模块

git submodule foreach --quiet 'echo $name'

Extract 30-day commits per repo

提取每个仓库30天内的提交

git log --since="30 days ago" --pretty=format:"%H|%s|%an|%ad" --date=short


**Data Collected:**
- Commit hash, message, author, date
- Files changed per commit
- Diff summaries
- Commit frequency patterns

git log --since="30 days ago" --pretty=format:"%H|%s|%an|%ad" --date=short


**收集的数据：**
- 提交哈希、提交信息、作者、日期
- 每次提交变更的文件
- Diff摘要
- 提交频率模式

2. ABSTRACT - Identify Patterns

2. ABSTRACT - 识别模式

Analyze collected data to identify recurring patterns:

Pattern Types:

Code Patterns: Import conventions, code structures, techniques
Workflow Patterns: TDD adoption, config-before-code, test-with-feature
Commit Patterns: Message conventions, prefixes (feat, fix, chore)
Correction Patterns: Fix commits, "actually" messages, immediate follow-ups
Tool Patterns: Framework usage, library adoption, tooling preferences

Pattern Detection Heuristics:

Frequency: Pattern appears in 3+ commits
Consistency: Same pattern used by multiple authors
Spread: Pattern appears across multiple repositories

分析收集的数据以识别重复出现的模式：

模式类型：

代码模式：导入规范、代码结构、技术实现
工作流模式：TDD实践、先配置后编码、随功能写测试
提交模式：提交信息规范、前缀（feat, fix, chore）
修正模式：修复提交、包含"actually"的提交信息、即时后续提交
工具模式：框架使用、库选型、工具偏好

模式检测规则：

频率：模式出现在3次以上提交中
一致性：多个作者使用相同模式
覆盖范围：模式出现在多个仓库中

3. GENERALIZE - Determine Scope

3. GENERALIZE - 确定范围

Categorize patterns by their applicability:

Scope	Criteria	Storage Location
Global	5+ repos	`~/.claude/memory/patterns/`
Domain	2-4 repos, same domain	`~/.claude/memory/domains/<domain>/`
Project	Single repo	`<repo>/.claude/knowledge/`

根据适用性对模式进行分类：

范围	判定标准	存储位置
全局	5个以上仓库使用	`~/.claude/memory/patterns/`
领域	2-4个仓库使用，同属一个领域	`~/.claude/memory/domains/<domain>/`
项目	单一仓库使用	`<repo>/.claude/knowledge/`

4. STORE - Persist and Act

4. STORE - 持久化并执行操作

Score patterns and take appropriate action:

Scoring Criteria:

Frequency (0.0-1.0): How often the pattern appears
Cross-repo Impact (0.0-1.0): How many repos use it
Complexity (0.0-1.0): Pattern sophistication
Time Savings (0.0-1.0): Estimated automation benefit

Weighted Score Calculation:

score = (frequency * 0.3) + (cross_repo * 0.3) + (complexity * 0.2) + (time_savings * 0.2)

Actions by Score:

Score Range	Action
>= 0.8	Create new skill automatically
0.6 - 0.79	Enhance existing skill
< 0.6	Log for future reference

为模式评分并执行相应操作：

评分标准：

频率（0.0-1.0）：模式出现的频次
跨仓库影响（0.0-1.0）：使用该模式的仓库数量
复杂度（0.0-1.0）：模式的复杂程度
时间节省（0.0-1.0）：自动化带来的预估时间收益

加权分数计算：

score = (frequency * 0.3) + (cross_repo * 0.3) + (complexity * 0.2) + (time_savings * 0.2)

分数对应的操作：

分数范围	操作
>= 0.8	自动创建新Skill
0.6 - 0.79	优化现有Skill
< 0.6	记录以便后续参考

State Management

状态管理

State File:

~/.claude/state/reflect-state.yaml

yaml

version: "1.0"
last_run: 2026-01-21T10:30:00Z
analysis_window_days: 30
repos_analyzed: 26
patterns_extracted: 45
actions_taken:
  skills_enhanced: 5
  skills_created: 2
  learnings_stored: 23
next_scheduled: 2026-02-21
history:
  - date: 2026-01-21
    patterns: 45
    skills_created: 2
    skills_enhanced: 5

状态文件：

~/.claude/state/reflect-state.yaml

yaml

version: "1.0"
last_run: 2026-01-21T10:30:00Z
analysis_window_days: 30
repos_analyzed: 26
patterns_extracted: 45
actions_taken:
  skills_enhanced: 5
  skills_created: 2
  learnings_stored: 23
next_scheduled: 2026-02-21
history:
  - date: 2026-01-21
    patterns: 45
    skills_created: 2
    skills_enhanced: 5

Pattern Output Format

模式输出格式

yaml

patterns:
  - id: "pattern-001"
    type: "workflow"
    name: "TDD Test-First Pattern"
    description: "Tests created before implementation"
    evidence:
      - repo: "aceengineer-admin"
        commits: ["abc123", "def456"]
      - repo: "digitalmodel"
        commits: ["ghi789"]
    frequency: 0.85
    cross_repo_score: 0.9
    complexity_score: 0.7
    time_savings_score: 0.8
    final_score: 0.83
    recommended_action: "create_skill"

yaml

patterns:
  - id: "pattern-001"
    type: "workflow"
    name: "TDD测试先行模式"
    description: "在实现功能前先编写测试"
    evidence:
      - repo: "aceengineer-admin"
        commits: ["abc123", "def456"]
      - repo: "digitalmodel"
        commits: ["ghi789"]
    frequency: 0.85
    cross_repo_score: 0.9
    complexity_score: 0.7
    time_savings_score: 0.8
    final_score: 0.83
    recommended_action: "create_skill"

Integration Points

集成点

With skill-learner

与skill-learner集成

Shares pattern extraction logic
Extended for multi-repo analysis
Complementary triggers (post-commit vs periodic)

共享模式提取逻辑
扩展为多仓库分析
互补触发方式（提交后 vs 周期性）

With repo-sync

与repo-sync集成

Uses parallel git operations
Leverages submodule enumeration

使用并行Git操作
利用子模块枚举功能

With skill-creator

与skill-creator集成

Invoked when score >= 0.8
Passes pattern data for skill generation

当分数≥0.8时触发
传递模式数据用于Skill生成

State Files Updated

更新的状态文件

```
~/.claude/state/reflect-state.yaml
```
: Reflection history
```
~/.claude/state/skills-progress.yaml
```
: Skill updates
```
.claude/skill-registry.yaml
```
: New skill entries

```
~/.claude/state/reflect-state.yaml
```
：复盘历史
```
~/.claude/state/skills-progress.yaml
```
：Skill更新记录
```
.claude/skill-registry.yaml
```
：新Skill条目

Execution Checklist

执行检查清单

Verify git access to all submodules (
```
git submodule status
```
)
Ensure state directory exists (
```
~/.claude/state/
```
)
Run with
```
--dry-run
```
first to preview patterns
Review extracted patterns before skill creation
Verify created/enhanced skills work correctly
Check state file for accurate tracking

验证所有子模块的Git访问权限（
```
git submodule status
```
）
确保状态目录存在（
```
~/.claude/state/
```
）
先使用
```
--dry-run
```
预览模式
在创建Skill前检查提取的模式
验证创建/优化后的Skill可正常工作
检查状态文件的跟踪是否准确

Error Handling

错误处理

Submodule Access Issues

子模块访问问题

bash

undefined

bash

undefined

Check submodule status

检查子模块状态

git submodule status

Update submodules

更新子模块

git submodule update --init --recursive

undefined

git submodule update --init --recursive

undefined

Empty History

无历史记录

If no commits found in window, reflection completes with warning:

Warning: No commits found in the last 30 days
Consider running with --days 90 for a larger window

如果在指定窗口内未找到任何提交，复盘会带着警告完成：

Warning: No commits found in the last 30 days
Consider running with --days 90 for a larger window

Pattern Scoring Issues

模式评分问题

If pattern scores seem incorrect:

Check evidence commit counts
Verify cross-repo detection
Review pattern categorization

如果模式分数看起来不准确：

检查提交证据的数量
验证跨仓库检测结果
检查模式分类是否正确

Workflows

工作流

Weekly Reflection

每周复盘

bash

undefined

bash

undefined

Quick weekly review

快速每周回顾

/reflect --days 7

Review patterns

查看模式

cat ~/.claude/state/reflect-state.yaml

undefined

cat ~/.claude/state/reflect-state.yaml

undefined

Monthly Deep Reflection

月度深度复盘

bash

undefined

bash

undefined

Full 30-day analysis

完整30天分析

/reflect

Extended with skill creation

扩展分析并创建Skill

/reflect --days 30

undefined

/reflect --days 30

undefined

Quarterly Review

季度回顾

bash

undefined

bash

undefined

Extended quarterly analysis

扩展季度分析

/reflect --days 90

Review all created skills

查看所有创建的Skill

ls .claude/skills/

undefined

ls .claude/skills/

undefined

Metrics & Success Criteria

指标与成功标准

Analysis Coverage: 100% of active submodules analyzed
Pattern Detection Rate: >= 5 patterns per reflection
Skill Creation Quality: Created skills rated useful by user
State Persistence: All runs tracked in state file
Performance: Full reflection completes in < 10 minutes

分析覆盖率：100%活跃子模块被分析
模式检测率：每次复盘至少检测到5个模式
Skill创建质量：创建的Skill被用户评为实用
状态持久化：所有运行记录都保存在状态文件中
性能：完整复盘在10分钟内完成

Best Practices

最佳实践

Run Frequency

运行频率

Weekly:
```
--days 7
```
for quick insights
Monthly: Default 30-day for comprehensive analysis
Quarterly:
```
--days 90
```
for strategic patterns

每周：使用
```
--days 7
```
获取快速洞察
每月：默认30天窗口进行全面分析
每季度：使用
```
--days 90
```
识别战略模式

Pattern Review

模式审查

Always use
```
--dry-run
```
before creating skills
Review high-scoring patterns manually
Verify cross-repo patterns are genuine

在创建Skill前始终使用
```
--dry-run
```
手动审查高分模式
验证跨仓库模式是否真实有效

Skill Creation

Skill创建

Check created skills compile/work
Add examples from actual commits
Link to source evidence

检查创建的Skill是否可编译/运行
添加来自实际提交的示例
链接到源证据

Scripts Architecture

脚本架构

The skill uses a modular pipeline of scripts:

daily-reflect.sh (orchestrator)
├── analyze-history.sh    # REFLECT: Extract git commits
├── extract-patterns.sh   # ABSTRACT: Identify patterns
├── analyze-trends.sh     # GENERALIZE: Cross-day trends
├── create-skills.sh      # STORE: Auto-create skills
└── generate-report.sh    # Weekly digest reports

该Skill使用模块化的脚本管道：

daily-reflect.sh (编排器)
├── analyze-history.sh    # REFLECT: 提取Git提交
├── extract-patterns.sh   # ABSTRACT: 识别模式
├── analyze-trends.sh     # GENERALIZE: 跨时间趋势分析
├── create-skills.sh      # STORE: 自动创建Skill
└── generate-report.sh    # 每周摘要报告生成

Script Details

脚本详情

Script	Phase	Input	Output
`analyze-history.sh`	REFLECT	Git repos	`analysis_*.json`
`extract-patterns.sh`	ABSTRACT	Analysis JSON	`patterns_*.json`
`analyze-trends.sh`	GENERALIZE	Multiple patterns	`trends_*.json`
`create-skills.sh`	STORE	Patterns	Skills + learnings
`generate-report.sh`	Report	All data	`weekly_digest_*.md`

脚本	阶段	输入	输出
`analyze-history.sh`	REFLECT	Git仓库	`analysis_*.json`
`extract-patterns.sh`	ABSTRACT	分析JSON文件	`patterns_*.json`
`analyze-trends.sh`	GENERALIZE	多个模式文件	`trends_*.json`
`create-skills.sh`	STORE	模式文件	Skill + 学习成果
`generate-report.sh`	报告生成	所有数据	`weekly_digest_*.md`

Output Locations

输出位置

~/.claude/state/
├── reflect-state.yaml       # Current state
├── reflect-history/         # Raw analysis files
│   └── analysis_*.json
├── patterns/                # Extracted patterns
│   └── patterns_*.json
├── trends/                  # Trend analysis
│   └── trends_*.json
└── reports/                 # Weekly digests
    └── weekly_digest_*.md

~/.claude/memory/patterns/
└── learnings.yaml           # Low-score patterns for reference

.claude/skills/workspace-hub/auto-generated/
└── <skill-name>/            # Auto-created skills
    └── SKILL.md

~/.claude/state/
├── reflect-state.yaml       # 当前状态
├── reflect-history/         # 原始分析文件
│   └── analysis_*.json
├── patterns/                # 提取的模式
│   └── patterns_*.json
├── trends/                  # 趋势分析
│   └── trends_*.json
└── reports/                 # 每周摘要
    └── weekly_digest_*.md

~/.claude/memory/patterns/
└── learnings.yaml           # 低分模式参考

.claude/skills/workspace-hub/auto-generated/
└── <skill-name>/            # 自动创建的Skill
    └── SKILL.md

References

参考链接

Skill Learner - Post-commit pattern extraction
Repo Sync - Multi-repo operations
Skill Creator - Skill generation

Skill Learner - 提交后模式提取
Repo Sync - 多仓库操作
Skill Creator - Skill生成

Version History

版本历史

2.0.0 (2026-01-21): Full RAGS loop implementation
- Added pattern extraction engine (
```
extract-patterns.sh
```
  )
- Added cross-daily trend analysis (
```
analyze-trends.sh
```
  )
- Added actionable reports generator (
```
generate-report.sh
```
  )
- Added auto-skill creation module (
```
create-skills.sh
```
  )
- Updated
```
daily-reflect.sh
```
  to orchestrate all phases
- Weekly reports auto-generated on Sundays
1.0.0 (2026-01-21): Initial release with basic RAGS spec

2.0.0 (2026-01-21): 完整RAGS循环实现
- 新增模式提取引擎（
```
extract-patterns.sh
```
  ）
- 新增跨时间趋势分析（
```
analyze-trends.sh
```
  ）
- 新增可执行报告生成器（
```
generate-report.sh
```
  ）
- 新增自动Skill创建模块（
```
create-skills.sh
```
  ）
- 更新
```
daily-reflect.sh
```
  以编排所有阶段
- 周日自动生成周报
1.0.0 (2026-01-21): 初始版本，包含基础RAGS规范