home / skills / project-n-e-k-o / n.e.k.o / post-work-analysis

post-work-analysis skill

safe

This skill archives post-work learnings by analyzing complex tasks and creating reusable insights for future debugging and development.

npx playbooks add skill project-n-e-k-o/n.e.k.o --skill post-work-analysis

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

2.5 KB

---
name: 📝 任务完成后归档
description: |
  **重要提醒**: 每次完成复杂调试或开发任务后，主动执行此流程！
  将学到的经验归档为 skill，供以后参考。不要等用户提醒。
---

# 📝 任务完成后归档 (Post-Work Analysis)

> **🔔 触发条件**: 当你刚完成一个涉及调试、研究、或多次迭代的任务时，主动执行此流程！

## 应用场景

- ✅ 解决了非平凡的 bug
- ✅ 通过调查发现了根本原因
- ✅ 解决方案涉及多次尝试
- ✅ 知识对未来类似问题有价值

## 归档流程

### 1. 识别直接原因
```
直接原因: [一句话描述]
例: "边界检测使用模型中心点而非可见区域，导致放大后只能看到腿"
```

### 2. 分析根本原因
深入思考 - 为什么会存在这个问题？
- **设计缺陷**: 架构问题？
- **工具/平台限制**: 已知 bug？缺失功能？
- **人为因素**: 开发中常见错误？
- **文档缺失**: 文档不清晰？

### 3. 判断是否值得归档为 Skill

| 情况 | 是否归档 |
|------|---------|
| 通用算法/公式（可复用于其他项目） | ✅ 归档 |
| 特定框架/库的陷阱和解决方案 | ✅ 归档 |
| 具体项目的代码组织问题 | ❌ 不归档 |
| 简单的拼写错误或配置问题 | ❌ 不归档 |

### 4. 创建/更新 Skill 文件

路径: `.agent/skills/<topic>/SKILL.md`

```markdown
---
name: [skill-name]
description: [简要描述这个 skill 涵盖的内容]
---

# [标题]

## 症状
- [可观察的问题 1]
- [可观察的问题 2]

## 根本原因
### 原因 1: [名称]
- **问题**: [描述]
- **为什么发生**: [根因分析]
- **诊断方法**: [如何确认]
- **解决方案**: [如何修复]

## 代码解决方案
[可复用的代码片段和公式]

## 关键经验
- [要点]
```

### 5. 向用户汇报
- 修复了什么
- 根因总结
- 知识归档位置
- 遗留问题（如有）

## 归档示例

### 示例 1: VRM 物理调试
- **直接原因**: 头部碰撞体过大
- **根因**: 模型缩放不一致、Unity 导出 bug
- **归档**: `.agent/skills/vrm-physics/SKILL.md`

### 示例 2: 3D 拖拽交互
- **直接原因**: 使用固定 panSpeed 导致缩放后移动不同步
- **根因**: 未考虑相机距离对像素→世界空间映射的影响
- **归档**: `.agent/skills/3d-interaction/SKILL.md`（可选）

## 提醒

> 🚨 **不要等用户提醒！** 每次完成复杂任务后，主动问自己：
> 1. 这个问题有通用价值吗？
> 2. 以后可能再遇到类似问题吗？
> 3. 如果是，立即归档！

Overview

This skill captures lessons after finishing complex debugging, research, or multi-iteration development work and turns them into reusable knowledge. It teaches how to identify direct and root causes, decide if results are worth archiving, create a concise skill entry, and report outcomes to stakeholders. Use it proactively every time you finish a non-trivial task.

How this skill works

After a task completes, the skill guides you to record the direct cause, perform root-cause analysis across design, tools, human factors, and documentation, then evaluate whether the finding is reusable. If worthwhile, it instructs how to create or update a formal skill entry in the agent knowledge base and how to summarize the fix for users. It ends by prompting follow-up and noting any remaining issues.

When to use it

After resolving a non-trivial bug that required investigation
When a solution involved multiple attempts or novel approaches
After discovering the root cause of recurring failures
When findings are likely useful for future projects or teammates
Whenever documentation or process gaps are uncovered

Best practices

Write one clear sentence for the direct cause to make search and indexing effective
Distinguish symptom, direct cause, and root cause—ask why repeatedly
Prefer archiving reusable algorithms, framework-specific pitfalls, and diagnostic patterns
Skip trivial fixes (typos, simple config) unless they reveal a deeper pattern
Include diagnostic steps and copy-paste-ready code snippets when applicable

Example use cases

Fixed a rendering bug caused by using model center instead of visible bounds and saved the detection pattern
Tracked down a physics glitch to inconsistent model scaling and archived the diagnostic checklist
Documented camera-distance effects on pan/zoom math that prevented repeated regressions
Created a library note about a framework export bug to warn future integrators
Summarized a multi-step workaround so support can apply it without re-investigation

FAQ

How do I decide if something is worth archiving?

If the solution is reusable beyond a single file, addresses framework/library pitfalls, or documents a diagnostic pattern, archive it. Skip one-off code organization or trivial typos.

What key sections should a skill entry contain?

Include symptoms, root-cause analysis, diagnostic methods, the fix or code snippet, and concise key takeaways so others can reproduce the diagnosis and solution.