This skill improves rough AI-generated output into refined, high-quality responses.
Overview
This skill improves AI's initial output from multiple angles and converts it into a more refined response. It enhances fluency, professionalism, clarity, structure, and readability, and optimizes it for the target audience and context.
Main Features
- Improving sentence fluency: Improve into natural, easy-to-read text
- Structuring: Add logical organization and headings
- Clarification: Make ambiguous expressions concrete
- Elaboration: Supplement missing information
- Conciseness: Remove redundant parts
- Tone adjustment: Adjust to formal, casual, technical, etc.
- Expertise enhancement: Improve technical accuracy and depth
- Readability improvement: Paragraph splitting, bullet points, adding code examples
- Adding examples: Suggestions for concrete examples, code snippets, diagrams
- Target audience optimization: Adjust for beginners, intermediate, advanced users
Improvement Patterns
1. Improving sentence fluency
Before: Awkward sentences
AI初期出力:
「Reactです。コンポーネントベースです。再利用できます。
JavaScriptライブラリです。UIを作ります。」
After: Fluent sentences
改善版:
「Reactは、再利用可能なコンポーネントベースでUIを構築できる
JavaScriptライブラリです。各コンポーネントは独立した機能を持ち、
複雑なユーザーインターフェースを効率的に開発できます。」
Improvement points:
- Combine short sentences into a natural flow
- Appropriate use of conjunctions
- Improve sentence rhythm and readability
2. Structuring
Before: Unstructured
AI初期出力:
「TypeScriptは型があります。JavaScriptのスーパーセットです。
コンパイルが必要です。エラーを早期に発見できます。大規模開発に向いています。
IDEのサポートが良いです。」
After: Structured
改善版:
# TypeScriptの特徴
## 概要
TypeScriptは、JavaScriptにオプショナルな型システムを追加した
スーパーセット言語です。
## 主な特徴
### 1. 静的型付け
- コンパイル時に型チェックを実施
- 実行前にエラーを発見可能
### 2. 開発体験の向上
- IDEによる強力な補完機能
- リファクタリングの安全性向上
### 3. スケーラビリティ
- 大規模プロジェクトに最適
- チーム開発での生産性向上
Improvement points:
- Hierarchical structure with headings
- Organize by topic
- Clarify key points with bullet points
3. Clarification
Before: Ambiguous
AI初期出力:
「このメソッドはデータを処理します。いくつかの変換を行い、
結果を返します。」
After: Clear
改善版:
「この`processData()`メソッドは、以下の処理を実行します:
1. 入力データのバリデーション
2. 数値データの正規化(0-1の範囲に変換)
3. 欠損値の補完(平均値で埋める)
4. 処理済みデータの配列を返却
**入力**: 数値の配列
**出力**: 正規化された数値の配列
**例外**: データが空の場合、`EmptyDataError`をスロー
Improvement points:
- Specify concrete processing details
- Clarify inputs and outputs
- Also describe exception cases
4. Elaboration
Before: Incomplete
AI初期出力:
「Dockerを使ってください。」
After: Detailed
改善版:
# Dockerの導入手順
## 1. Dockerのインストール
### Windows
1. Docker Desktopをダウンロード
2. インストーラーを実行
3. WSL 2を有効化(Windows 10/11の場合)
### macOS
```bash
brew install --cask docker
Linux (Ubuntu)
sudo apt update
sudo apt install docker.io
sudo systemctl start docker
sudo systemctl enable docker
2. Creating a Docker image
Create a Dockerfile:
FROM node:18-alpine
WORKDIR /app
COPY package*.json ./
RUN npm ci
COPY . .
CMD ["npm", "start"]
3. Building and running the image
# イメージビルド
docker build -t myapp:1.0 .
# コンテナ起動
docker run -p 3000:3000 myapp:1.0
Troubleshooting
- Port is in use:
-pSpecify a different port with - Permission error:
sudoAdd, or add the user to the docker group
**改善ポイント**:
- 具体的な手順を追加
- コード例の提供
- トラブルシューティング情報を含める
### 5. 簡潔化
#### Before: 冗長
AI initial output: "This algorithm first scans each element of the array in order. Then, during the scanning process, it checks each element to see whether that element satisfies the condition. If it satisfies the condition, it adds that element to a new array. It repeats this process until the last element of the array. Ultimately, a new array containing only the elements that satisfied the condition is obtained."
#### After: 簡潔
Improved version: "This algorithm scans each element of the array, extracts only the elements that satisfy the condition, and generates a new array.
const filtered = array.filter(element => element > 10);
In other words,Array.filter()This is equivalent to the method."
**改善ポイント**:
- 冗長な説明を削除
- コード例で直感的に理解
- 1-2文に要約
### 6. トーン調整
#### フォーマル(ビジネス文書)
AI initial output (casual): "I found a bug, so I fixed it. Tests are OK too."
Improved version (formal): "We have identified the program defect and completed the fix. The validity of the fix has been confirmed through unit tests and integration tests."
#### カジュアル(ブログ記事)
AI initial output (stiff): "In this protocol, HTTP/2 is adopted in order to realize communication between the client and server."
Improved version (casual): "This app uses HTTP/2 for communication between the client and server. It's faster than the conventional HTTP/1.1!"
#### 技術的(ドキュメント)
AI initial output (general audience): "This program is fast."
Improved version (technical): "This algorithm operates with a time complexity of O(n log n), and demonstrates excellent performance in the average case. The space complexity is O(log n), achieving efficiency close to in-place sorting."
### 7. 専門性向上
#### Before: 表面的
AI initial output: "React is fast because it uses a virtual DOM."
#### After: 専門的
Improved version:
Performance optimization by React's virtual DOMHow the virtual DOM works
React keeps a lightweight DOM representation called the virtual DOM (Virtual DOM) in memory. When the state changes, it updates efficiently through the following process:
-
Reconciliation (diff detection)
- Build a new virtual DOM tree
- Compare with the previous virtual DOM tree (diffing algorithm)
- Identify the minimum nodes that need to be changed
-
Batch Update
- Combine multiple state changes into a single DOM operation
- Reduce the number of layout recalculations
-
Commit Phase
- Apply changes to the actual DOM
- Trigger browser re-rendering
Performance advantages
- Minimization of DOM operations: Diff detection from O(n³) to O(n)
- Batch processing: Execute multiple updates at once
- Asynchronous rendering: React 18のConcurrent Rendering
Notes
The virtual DOM is not always fast:
- For small changes, overhead may occur
- Appropriate memoization (
React.memo,useMemo) is important
**改善ポイント**:
- 技術的な詳細を追加
- アルゴリズムの説明
- 計算量の明記
- 制約・注意点も記載
### 8. 読みやすさ改善
#### Before: 読みにくい
AI initial output: "Error handling uses try-catch. In the catch block, catch the error, output it to the log, and display a message to the user. In the finally block, clean up resources."
#### After: 読みやすい
Improved version:
Error Handling Best PracticesBasic structure
try {
// エラーが発生する可能性のある処理
const data = await fetchData();
processData(data);
} catch (error) {
// エラーハンドリング
console.error('エラー:', error);
showErrorMessage('データの取得に失敗しました');
} finally {
// リソースのクリーンアップ(常に実行)
closeConnection();
}
Execution flow
- Try Block: Execute normal processing
- Catch Block: Execute when an error occurs
- Record error log
- Notify the user
- Finally Block: Execute regardless of success or failure
- Release resources
- Post-processing
**改善ポイント**:
- コード例で視覚的に理解
- ステップバイステップの説明
- 見出しで構造化
### 9. 例の追加
#### Before: 抽象的
AI initial output: "A Promise is an object for handling asynchronous processing."
#### After: 具体例付き
Improved version:
Introduction to JavaScript PromisesOverview
A Promise is an object for handling asynchronous processing, and it represents the completion or failure of a process.
Basic usage
Example 1: API call
// Promiseを返す関数
function fetchUser(userId) {
return fetch(`/api/users/${userId}`)
.then(response => response.json())
.then(data => data.user);
}
// 使用例
fetchUser(123)
.then(user => console.log('ユーザー:', user))
.catch(error => console.error('エラー:', error));
Example 2: Multiple asynchronous processes
// 並列実行
Promise.all([
fetchUser(1),
fetchUser(2),
fetchUser(3)
])
.then(users => {
console.log('全ユーザー:', users);
});
// 順次実行
async function processUsers() {
const user1 = await fetchUser(1);
const user2 = await fetchUser(2);
console.log(user1, user2);
}
Example 3: Error handling
fetchUser(999)
.then(user => {
if (!user) {
throw new Error('ユーザーが見つかりません');
}
return user;
})
.catch(error => {
console.error('処理失敗:', error.message);
return null; // デフォルト値
});
Common Patterns
| Patterns | Use case | Example |
|---|---|---|
| Promise.all | Parallel execution (all succeed) | Simultaneous calls to multiple APIs |
| Promise.race | Adopt the fastest result | Timeout handling |
| Promise.allSettled | Parallel execution (regardless of success/failure) | Best-effort processing |
**改善ポイント**:
- 複数の具体例
- ユースケース別の説明
- 表による比較
### 10. 対象読者最適化
#### 初心者向け
"React Hooks are a convenient feature that allows state management in function components.
The most commonly used Hook: useState
import { useState } from 'react';
function Counter() {
// カウンターの値を保持する
const [count, setCount] = useState(0);
return (
<div>
<p>現在のカウント: {count}</p>
<button onClick={() => setCount(count + 1)}>
+1
</button>
</div>
);
}
Key Points:
useState(0)Set the initial value to 0 withcountGet the current value withsetCount()Update the value with
#### 上級者向け
"React Hooks Internal Implementation and Performance Optimization"
How Hooks are implemented
React internally uses the Fiber architecture and maintains a linked list of hooks for each component instance.
// 簡略化した内部表現
type Hook = {
memoizedState: any;
next: Hook | null;
};
// Fiber node
type Fiber = {
memoizedState: Hook | null;
// ...
};
Performance optimization
1. Memoization with useCallback
const memoizedCallback = useCallback(
() => {
doSomething(a, b);
},
[a, b] // 依存配列
);
Internally, it returns the same function reference as long as the dependency array does not change.
2. Optimization points for useMemo
const expensiveValue = useMemo(() => {
return computeExpensiveValue(a, b);
}, [a, b]);
Note: Memoization itself also has a cost, so it is recommended to use it only in the following cases:
- The computation cost is high (>50ms)
- Props to child components (reference consistency is required)
## 改善プロセス
### ステップ1: 初期分析
Input: AI initial output
Analysis items:
- Fluency of the text
- Presence of structure
- Completeness of information
- Clarity of the target audience
- Level of expertise
Output: List of improvement points
### ステップ2: 改善計画
Prioritization:
- Correction of major errors/deficiencies
- Structuring
- Clarification/detailing
- Refinement of the text
- Addition of examples
### ステップ3: 改善実行
Apply each improvement item:
- Rewriting the text
- Adding headings
- Inserting code examples
- Supplementing explanations
### ステップ4: 品質確認
Checklist: ✓ Accuracy of information ✓ Logical flow ✓ Readability ✓ Suitability for the target audience ✓ Completeness
## 使用例
### 基本的な改善
Please improve the following AI output:
"React is convenient. It is easy to use. It is popular."
Improvement policy:
- Make the sentences fluent
- Add specific reasons
- Structure it
### トーン調整
Please rewrite the following technical explanation in a casual blog-article style:
"In this system, by adopting a microservices architecture, we have achieved improved scalability."
Target audience: General developers Tone: Friendly, approachable
### 対象読者最適化
Please rewrite the following explanation for beginners:
"async/await is syntactic sugar for Promises, and it makes it possible to write asynchronous processing like sequential code."
Target: Programming beginners (understands JavaScript basics) Improvement policy:
- Make technical terms plain
- Add concrete examples
- Step-by-step explanation
### 詳細化
Please elaborate on the following incomplete explanation:
"Create a Dockerfile and build the image."
Content to add:
- The specific contents of the Dockerfile
- Build Command
- How to run it
- Troubleshooting
## ベストプラクティス
1. **原文の意図を保持**: 改善しても元の意味を変えない
2. **過度な装飾を避ける**: 必要な情報を明確に
3. **対象読者を常に意識**: 知識レベルに合わせた表現
4. **具体例を活用**: 抽象的な説明には例を追加
5. **構造化**: 見出し、箇条書きで読みやすく
6. **バランス**: 詳細すぎず、簡潔すぎず
7. **一貫性**: トーン、用語の使用を統一
## バージョン情報
- スキルバージョン: 1.0.0
- 最終更新: 2025-11-22
---
## 使用例まとめ
### シンプルな改善
Please improve this sentence: [rough output]
### 詳細な改善
Please improve the following AI output: [text]
Improvement policy:
- Fluency of the sentences
- Structure
- Elaboration
- Tone: Formal
- Target: Beginners
このスキルで、AI出力を洗練された高品質な応答に変換しましょう!