Work Log Analyzer

Purpose

This Skill analyzes work logs, journals, and development notes to help developers:

1. Track project evolution - Understand how implementations, features, or decisions evolved over time
2. Manage TODOs - Find pending, in-progress, and overdue tasks from logs
3. Extract insights - Discover patterns, decisions, and rationale behind implementations
4. Answer questions - Query logs with natural language to find specific information

The Skill helps developers:

• Quickly find information in lengthy work logs
• Track decision evolution and rationale
• Manage TODOs without external tools
• Generate status reports from logs
• Maintain project continuity

---

Instructions

When the user provides a log file and asks a question (for example: "關於 SellerCheck 實作方案的演進", "我還有哪些 TODO 還沒完成？", "Lite Engagement 在何時才把 L10n 翻譯確定"), follow the steps below.

Execution Steps

1. Determine the skill base directory from the skill invocation context (provided when the skill is loaded).

   Important: When using the aggregation script, always use the full absolute path:
   <skill_base_directory>/scripts/aggregate_activities.sh, not a relative path, since the
   current working directory may be the user's project directory.

2. Identify the log file

   • User provides explicit file path: use it directly
   • User says "my work log" or "my journal": ask for the file path
   • User pastes content directly: analyze the provided text

3. Read and parse the log

   • Read the file content using the available file reading tool
   • Parse the structure (dates, sections, TODOs, etc.)
   • Identify the format (Markdown, plain text, structured, unstructured)

4. Understand the query — classify into one of 5 query types:

   • Activity Aggregation (活動聚合): "aggregate my activities", "show me today's work", "what did I do this week"

     • Use <skill_base_directory>/scripts/aggregate_activities.sh to aggregate structured activity records from activity-logger
     • Filter by date range (today, yesterday, this-week, last-week, this-month, all), project, type, or tag
     • Output modes: by-date (chronological), by-project, by-type, json (raw)

   • Timeline (時間軸): "X 的演進", "how did X evolve", "X 的變化"

     • Extract all entries related to the topic, sort chronologically
     • Summarize major milestones and decisions
     • Also query activity records for technical evolution

   • TODO Management (TODO 管理): "未完成的 TODO", "過期的任務", "pending tasks"

     • Extract all TODO items, categorize by status (pending, in-progress, completed)
     • Use current system date (or latest log date as fallback) to calculate overdue tasks
     • Resolve relative dates (e.g., "yesterday", "last week") and prioritize overdue items

   • Decision Tracking (決策追溯): "為什麼選擇 X", "X 的決策過程", "when was X decided"

     • Find relevant discussions, extract rationale and context
     • Identify decision makers or influencing factors
     • Also search activity contexts for decision rationale

   • General Search (通用搜尋): keyword or topic search

     • Find all mentions, provide surrounding context, group related entries

5. Analyze and synthesize

   • Connect related entries across different dates
   • Identify patterns and trends
   • Highlight important context
   • Note any gaps or missing information

6. Present findings

   • Structure the answer clearly
   • Include dates and references
   • Quote relevant excerpts
   • Summarize key insights
   • Suggest follow-up questions if appropriate

Output Requirements

Your response should be:

• Clear and structured: Use headings, bullet points, and quotes
• Chronological when appropriate: Timeline queries should show progression
• Action-oriented for TODOs: Show status, due dates, and priorities. Always state the reference date used for calculating overdue tasks (e.g., "分析日期: 2026-01-13")
• Context-rich: Include surrounding information to aid understanding
• Referenced: Cite dates or log sections for traceability

---

Supported Log Formats

Recommended Format (Structured Markdown)

# 2026-01-13

## SellerCheck Implementation

- Decided to use PostgreSQL for seller verification data
- Considered Redis but concerned about persistence
- Next: prototype the schema

## TODOs

- [ ] TODO: Implement SellerCheck API (due: 2026-01-20) #high-priority
- [x] TODO: Fix L10n issues in Lite Engagement (completed: 2026-01-10)
- [ ] TODO: Code review for PR #123

## Decisions

**Decision**: Use PostgreSQL for SellerCheck
**Rationale**: Need ACID guarantees and complex queries
**Alternatives considered**: Redis, MongoDB
**Date**: 2026-01-13

Also Supports

• Plain text logs: Less structured but still analyzable
• Bullet point journals: Simple daily logs
• Unstructured notes: Will extract information best-effort
• Mixed formats: Adapts to varying styles

TODO Format Recognition

The Skill recognizes these TODO patterns:

- [ ] TODO: Task description
- [x] TODO: Completed task
- TODO: Task without checkbox
- [ ] Task (due: YYYY-MM-DD)
- [ ] Task #priority-tag
- FIXME: Code fix needed
- HACK: Technical debt item

Format Tolerance: The Skill attempts to be lenient with common formatting variations:

• Checkboxes with missing spaces: - [] or - [X] (though - [ ] and - [x] are preferred)
• Various completion markers: [x], [X], [✓], [✔]
• In-progress markers: [~], [→], [...]
• Different TODO keywords: TODO, FIXME, HACK, BUG, NOTE

For best results, maintain consistent formatting throughout your logs.

---

Constraints

• File size: Large files (>10,000 lines) may need to be split
• Date parsing: Works best with ISO format (YYYY-MM-DD) or clear date headers
• TODO detection: Most accurate with consistent formatting
• Context: Cannot access files or code references mentioned in logs (only the log content itself)
• No write operations: This Skill only reads and analyzes, does not modify logs

---

Examples

Example 1: Activity Aggregation

User:

Aggregate my activities from this week

Expected behavior:

1. Run <skill_base_directory>/scripts/aggregate_activities.sh -r this-week
2. Parse the Markdown output
3. Present a formatted summary of the week's activities

Example output:

# 本週工作摘要 (2026-01-08 ~ 2026-01-13)

## 完成的任務 ✅
- 實作 activity-logger skill (2026-01-13)
  - 跨 session 活動追蹤功能
  - Git 整合和自動 context 捕捉
  - 專案：agent-skills

## 修復的問題 🐛
- 修正 Git 路徑一致性 (2026-01-13)
- 移除 credentials 安全性問題 (2026-01-13)
- 優化 ARG_MAX 處理 (2026-01-13)

## 重構工作 ♻️
- 優化 jq 批次處理效能 (2026-01-13)

## 統計
- 總活動數：6
- 涉及專案：1 (agent-skills)
- 變更檔案：約 30 個

---

User:

Show me all bug fixes from last month

Expected behavior:

1. Run <skill_base_directory>/scripts/aggregate_activities.sh -r this-month -t bug_fixed
2. Filter activities by type "bug_fixed"
3. Present chronologically with context

---

User:

What did I work on in the security-project yesterday?

Expected behavior:

1. Run <skill_base_directory>/scripts/aggregate_activities.sh -r yesterday -p security-project
2. Show project-specific activities
3. Include file changes and tags

---

Example 2: Timeline

User:

我有一個 work.md 檔案，幫我分析「關於 SellerCheck 實作方案的演進」

Expected behavior:

1. Read the work.md file
2. Search for all entries mentioning "SellerCheck"
3. Extract entries chronologically
4. Summarize the evolution:
   • Initial consideration (date, initial thoughts)
   • Design iterations (what changed and why)
   • Final decision (current approach)
5. Present as a timeline with key milestones

Example output:

## SellerCheck 實作方案演進

### 2026-01-05: 初步討論
- 考慮使用 Redis 做快取層
- 擔心資料持久性問題

### 2026-01-08: 架構調整
- 決定改用 PostgreSQL 作為主要儲存
- 原因：需要 ACID 保證和複雜查詢能力
- Redis 仍用於快取層

### 2026-01-13: 最終方案
- 實作混合架構：PostgreSQL + Redis
- 已完成 schema 設計
- 下一步：實作 API endpoints

---

Example 3: TODO Management

User:

檢查我的 journal.md，哪些 TODO 還沒完成？哪些已經過期？

Expected behavior:

1. Read journal.md
2. Extract all TODO items
3. Parse status ([ ] vs [x])
4. Parse due dates if present
5. Compare with current date (2026-01-13)
6. Categorize and present

Example output:

## TODO 狀態報告

### ⚠️ 過期任務 (2 items)
1. **Implement SellerCheck API** (due: 2026-01-10)
   - Status: Not started
   - Priority: High
   - 來源: 2026-01-05 entry

2. **Review PR #456** (due: 2026-01-12)
   - Status: Not started

### 📋 待處理任務 (3 items)
1. **Fix L10n in Lite Engagement** (due: 2026-01-20)
   - Status: In progress
   - 最後更新: 2026-01-11

2. **Write unit tests for auth module**
   - No due date
   - 提到於: 2026-01-09

3. **Update documentation** (due: 2026-01-25)
   - Status: Not started

### ✅ 近期完成 (last 7 days)
1. **Setup CI/CD pipeline** (completed: 2026-01-08)
2. **Fix security vulnerability** (completed: 2026-01-12)

---

Example 4: Decision Tracking

User:

從 devlog.txt 找出「Lite Engagement 的 L10n 翻譯何時確定的？」

Expected behavior:

1. Read devlog.txt
2. Search for "Lite Engagement" AND "L10n" or "翻譯"
3. Find decision-related entries
4. Extract timeline and rationale
5. Present with context

Example output:

## Lite Engagement L10n 翻譯決策時間線

### 2026-01-03: 初步討論
> "開始考慮 Lite Engagement 的多語系支援，尚未確定翻譯範圍"

### 2026-01-08: 範圍確認
> "與 PM 確認 L10n 範圍：支援英文、中文（繁簡）、日文"
> "決定使用 i18next 作為翻譯框架"

### 2026-01-11: **最終確定** ✓
> "L10n 翻譯已經確定完成，所有字串已交付翻譯團隊"
> "預計 2026-01-20 收到翻譯結果"

**答案**: L10n 翻譯在 **2026-01-11** 確定完成並送出。

---

Example 5: General Search

User:

搜尋 notes.md 中關於「authentication」的所有提及

Expected behavior:

1. Read notes.md
2. Find all mentions of "authentication"
3. Extract surrounding context
4. Group by relevance or date
5. Present findings

---

Use Cases

Daily Standup Preparation

• Query: "昨天完成了什麼？" or "yesterday's completed tasks"
• Get a quick summary of recent work

Sprint Retrospective

• Query: "過去兩週的重要決策" or "decisions made in the last 2 weeks"
• Review and discuss team decisions

Technical Debt Tracking

• Query: "所有 HACK 和 FIXME 項目"
• Identify technical debt from logs

Decision Documentation

• Query: "為什麼選擇 X 架構？"
• Generate Architecture Decision Records from log entries

Onboarding New Team Members

• Query: "Y 功能的演進過程"
• Help new members understand project history

Personal Productivity

• Query: "這個月我完成了哪些任務？"
• Track personal accomplishments

---

Best Practices for Log Writing

To get the most from this Skill, consider these logging practices:

1. Use Consistent Date Headers

# 2026-01-13
## Morning Session

2. Structure TODOs Clearly

- [ ] TODO: Clear task description (due: 2026-01-20) #tag

3. Document Decisions Explicitly

## Decision: Use PostgreSQL
**Why**: Need ACID guarantees
**Alternatives**: Redis, MongoDB
**Trade-offs**: Slower than Redis, but more reliable

4. Use Contextual Tags

#SellerCheck #architecture #database

5. Link Related Entries

See 2026-01-10 entry for initial discussion

6. Regular Updates

• Update TODOs when status changes
• Add completion dates for finished tasks
• Document rationale for decisions

---

Advanced Queries

Combining Criteria

• "過期的 high-priority TODOs"
• "Last week's decisions about authentication"

Trend Analysis

• "How did the SellerCheck design change over time?"
• "What were the main concerns in January?"

Cross-Topic Analysis

• "All decisions related to database choices"
• "TODOs blocked by external dependencies"

Status Reports

• "Generate a weekly summary"
• "What's the current status of Project X?"

---

Tips for Users

1. Keep logs consistently: Regular logging makes analysis more valuable
2. Use clear dates: ISO format (YYYY-MM-DD) works best
3. Tag important entries: Use hashtags for easy filtering
4. Be specific in queries: More context = better results
5. Review TODO format: Consistent format improves detection
6. Include "why" not just "what": Decisions are more useful with rationale
7. Update completion status: Mark tasks as done to track progress
8. Split large logs: Consider monthly or quarterly files for better performance

---

Integration with Other Skills

• code-story-teller: Combine code history with work log context
• activity-logger: Structured activity records for aggregation

---

Limitations

• Cannot access external files referenced in logs
• No automatic log updating (read-only)
• Date parsing may fail on ambiguous formats
• Very unstructured logs may yield less precise results
• Cannot track TODOs across multiple files simultaneously (analyze one at a time)

---

Output Language

• Analysis and answers: Traditional Chinese (matches user's CLAUDE.md preference)
• Code snippets and technical terms: English
• Dates and times: ISO format or as they appear in logs
• Quotes from logs: Original language preserved

Want to turn this into a formal report? Try report-generator.