Daily Log Skill

Generate comprehensive daily operation logs to track work, decisions, and lessons learned.

When to Use

Use this skill at the end of a work session or day to:

• Record completed tasks and their outcomes
• Track token usage and time spent
• Document key decisions and their rationale
• Capture lessons learned and mistakes
• Maintain continuity across sessions

---

Log Format Templates

Template A: Full Detail (Legacy)

Use for: Important milestones, detailed project records See: FULL_TEMPLATE

Template B: Attention-Driven (Recommended)

Use for: Daily work logging, quick review See below ⬇️

---

Attention-Driven Log Format (v1.1)

# YYYY-MM-DD 操作日志

## 📅 会话概览
- **日期**: YYYY-MM-DD
- **工作时段**: HH:MM - HH:MM (X小时X分钟)
- **核心成果**: [一句话总结当天最重要的产出]
- **关键决策**: [X] 个
- **经验教训**: [X] 个
- **Token 消耗**: ~XX,XXX

---

## ⏱️ 时间分布

| 时段 | 任务 | 时长 | 注意力权重 |
|------|------|------|-----------|
| HH:MM-HH:MM | [任务1] | X分钟 | 9/10 |
| HH:MM-HH:MM | [任务2] | X分钟 | 7/10 |
| ... | ... | ... | ... |

**时间分析**:
- 高注意力任务耗时: X% (主要集中在XX:XX-XX:XX)
- 中断/切换次数: X 次
- 效率峰值时段: XX:XX-XX:XX

---

## 🎯 高注意力任务 (权重 8-10)

### [任务名称] (权重: X/10, 时段: HH:MM-HH:MM, 耗时: X分钟)

**一句话总结**: [核心成果或决策]

**关键细节**:
- [具体数据/数字]
- [文件路径/名称]
- [决策原因]
- [验证结果]

**经验教训** (如适用):
- [学到的要点]

---

## 📋 中注意力任务 (权重 5-7)

| 任务 | 权重 | 时段 | 关键成果 |
|------|------|------|----------|
| [任务名] | 7/10 | HH:MM-HH:MM | [一句话描述] |
| [任务名] | 6/10 | HH:MM-HH:MM | [一句话描述] |

---

## 📝 低注意力任务 (权重 0-4)

- [HH:MM-HH:MM] [任务名] - [状态]
- [HH:MM-HH:MM] [任务名] - [状态]

---

## 📊 今日统计

| 项目 | 数值 |
|------|------|
| 高注意力任务 | X |
| 中注意力任务 | X |
| 低注意力任务 | X |
| 代码文件创建 | X |
| 代码文件修改 | X |
| Skill 创建/更新 | X |
| Token 消耗 | ~XX,XXX |
| Git 提交 | X |

---

## 💡 今日最大教训

**一句话总结**: [核心教训]

**背景**: [发生了什么]
**根本原因**: [为什么发生]
**改进措施**: [如何改进]

---

## 🔗 关键文件位置

### 高价值产出
- `path/to/key/file1` - [一句话描述]
- `path/to/key/file2` - [一句话描述]

---

*日志生成时间: YYYY-MM-DD HH:MM*  
*注意力评分: 高[X] 中[X] 低[X]*

---

Attention Scoring System

How to Score Task Attention (0-10)

Factor	Weight	Indicator	Examples
关键决策	+3	改变了方向或方案	选择方案B、批准实施、确认规范
教训/错误	+3	发现问题并修复	违反规则、编译错误、逻辑bug
里程碑	+2	重要节点完成	MVP完成、发布上线、功能验收
文件变更	+1/个	创建/修改/删除文件	新建Skill、修改配置、重构代码
普通操作	0	常规查询或查看	查看状态、读取文件、检查日志

Attention Level Guidelines

Score 8-10 (High): 
  → Full detail: summary + key details + lessons
  
Score 5-7 (Medium): 
  → Brief: one sentence summary + key outcomes
  
Score 0-4 (Low): 
  → Minimal: title + status only

Examples

Task: "设计 MissionSystem 架构方案"

• 关键决策: +3 (选择了TK_SERIAL方案)
• 里程碑: +2 (设计完成)
• Score: 8/10 → High attention

Task: "修复编译错误"

• 教训: +3 (学会了BinaryReader→TK转换)
• 文件变更: +8个文件修改 = +1 (max)
• Score: 9/10 → High attention

Task: "查看 git status"

• 普通操作: 0
• Score: 2/10 → Low attention

---

Workflow

Step 1: Review Session

At end of session/day:

1. List all tasks completed
2. Identify major decisions made
3. Note any mistakes or lessons
4. Check for milestones reached

Step 2: Score Each Task

Apply attention scoring:

For each task:
  - Did it involve a key decision? (+3)
  - Was there a mistake/lesson? (+3)
  - Was it a milestone? (+2)
  - How many files changed? (+1 per, max 2)
  - Sum → Attention Score (0-10)

Step 3: Categorize by Attention Level

• High (8-10): Write detailed section
• Medium (5-7): Add to table
• Low (0-4): List as bullet points

Step 4: Extract Key Information

For high-attention tasks, extract:

• One-sentence summary
• Key details (numbers, paths, outcomes)
• Lessons learned (if applicable)

Step 5: Generate Log

Write to memory/YYYY-MM-DD.md using attention-driven template

Step 6: Update Long-term Memory (Optional)

If significant decisions or patterns emerged, update MEMORY.md

---

Best Practices

✅ Do

• Score honestly - Not every task is high attention
• Focus on value - What would you want to remember in a month?
• Quantify - Use numbers, file counts, token estimates
• Link key files - Only high-value outputs need paths
• One lesson max - Focus on the most important lesson of the day

❌ Don't

• Don't over-document low-attention tasks
• Don't skip lessons learned section
• Don't include full conversation transcripts
• Don't log routine checks (git status, etc.) unless relevant
• Don't wait too long (score while memory is fresh)

---

Comparison: Full Detail vs Attention-Driven

Scenario: MissionSystem MVP Implementation Day

Full Detail Version: ~500 lines, ~95,000 tokens to read

• Every task fully documented
• All file paths listed
• Complete error descriptions
• Full conversation context

Attention-Driven Version: ~150 lines, ~20,000 tokens to read

• 2-3 high-attention tasks detailed
• 3-4 medium tasks in table
• 5+ low tasks as bullets
• Key decisions and lessons highlighted

Review Time:

• Full Detail: 10-15 minutes to scan
• Attention-Driven: 2-3 minutes to understand

---

Version History

• v1.1 (2026-02-12) - Added Attention-Driven logging

  • Attention scoring system (0-10)
  • Three-level detail format
  • Focus on high-value information
  • Reduced log size by 60-70%

• v1.0 (2026-02-10) - Initial release

  • Standardized log format
  • 7-section structure
  • Statistics tracking
  • Lessons learned framework