REA
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Intelligent prompt improver hook for Claude Code. Type vibes, ship precision.
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/skills/severity1-claude-code-prompt-improver)在同一工作中把 AI 连到服务的 MCP
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Codebase intelligence for TypeScript and JavaScript, exposed as structured MCP analysis tools.
MCP server for terminal commands, file operations, and process management
同类工作,排序与 Skill 列表相同
| 本项 claude-code-prompt-improver | rea | graphify | archify | security-audit-skill | |
|---|---|---|---|---|---|
| 作者 | severity1 | morluto | Graphify-Labs | tt-a1i | cloudflare |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 1,972 | 461,751 | 148,352 | 96,757 | 73,156 |
| 最近代码更新 | 2026年10月1日 | 2026年11月16日 | 2026年11月16日 | 2026年11月15日 | 2026年9月15日 |
| 许可证 | MIT | MIT | Apache-2.0 | Apache-2.0 | MIT |
severity1 提供的 1 个 Skill(指令集),AI 读取后照做 · 不是 MCP server
MCP = 手
把 AI 连到外部服务或数据,例如读取 Notion 或开发票。需要有程序或服务器在运行。
Skill = 手册
教 AI 怎么做某项工作的文本文件,不连接任何服务。
有些会重叠:少数 Skill 带有调用工具的脚本,少数 MCP 只提供知识。请从你想完成的工作出发。
Claude Code
依据开发者 README 和 repo 中的文件夹
GitHub account severity1 (person), 14 years old
★ 1,972 stars · code last changed 2026年10月1日
No, plain text files the AI reads and follows
No scripts, so it does not go online by itself
Installing the maker’s way adds hooks that run without you asking: run every time you send a message · run before every tool the AI uses · run on the SubagentStart event
Checked automatically from the repo files at commit 50aae18 (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
第一步:你用哪个 AI?
选择你用的应用,本页只显示该应用的安装方法
Uploading to claude.ai would fail: the description is over 200 characters (226). Use "Let the AI install it" instead
复制这句话,粘贴到 Claude Code 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
重新打开 Claude Code,输入 /prompt-improver,或直接提出相关任务,Claude 会自动选用
Checked against official docs 2026年11月12日 · Source: code.claude.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
ChatGPT 桌面应用内置 Codex,Codex 从 ~/.agents/skills 文件夹读取 Skill
在 ChatGPT 桌面应用中打开 Codex,把这句话粘贴到对话中
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
把 prompt-improver 文件夹放到 ~/.agents/skills/prompt-improver
~ 是你的主文件夹($HOME)
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
不能使用你自己安装的 Skill,只能使用目录中插件自带的 Skill
想在 ChatGPT 中使用这个 Skill,请按上面的方法使用 ChatGPT 桌面应用
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
复制这句话,粘贴到 Codex 的对话中 · 适用于 ChatGPT 桌面应用中的 Codex、Codex CLI 和 IDE 扩展
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex · 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
复制这句话,粘贴到 Cursor 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
在 Cursor 的 Agent 聊天中输入 / 并选择 prompt-improver
Checked against official docs 2026年11月12日 · Source: cursor.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
适用于其他能读取 Skill 的应用。这句话列出了 Claude Code、Codex 和 Cursor 的路径,AI 会按你的应用选择 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026年11月12日 · Source: code.claude.com, learn.chatgpt.com, cursor.com
其他支持 Agent Skills 的应用(如 Gemini CLI、GitHub Copilot)使用不同文件夹,请查看该应用的指南,再把含 SKILL.md 的文件夹复制过去。
已于 2026年11月11日 对照 Claude Code、Codex 和 Cursor 官方指南核实 skill 文件夹
以下文字来自开发者的 repo,不是 PopNx MCP 的指令
6,643 个字符 · 在 GitHub 打开完整文件
---
name: prompt-improver
description: This skill enriches vague prompts with targeted research and clarification before execution. Should be used when a prompt is determined to be vague and requires systematic research, question generation, and execution guidance.
---
# Prompt Improver Skill
## Purpose
Transform vague, ambiguous prompts into actionable, well-defined requests through systematic research and targeted clarification. This skill is invoked when the hook has already determined a prompt needs enrichment.
## When This Skill is Invoked
**Automatic invocation:**
- UserPromptSubmit hook evaluates prompt
- Hook determines prompt is vague (missing specifics, context, or clear target)
- Hook invokes this skill to guide research and questioning
**Manual invocation:**
- To enrich a vague prompt with research-based questions
- When building or testing prompt evaluation systems
- When prompt lacks sufficient context even with conversation history
**Assumptions:**
- Prompt has already been identified as vague
- Evaluation phase is complete (done by hook)
- Proceed directly to research and clarification
## Core Workflow
This skill follows a 4-phase approach to prompt enrichment:
### Phase 1: Research
Create a dynamic research plan using TodoWrite before asking questions.
**Research Plan Template:**
1. **Check conversation history first** - Avoid redundant exploration if context already exists
2. **Review codebase** if needed:
- Task/Explore for architecture and project structure
- Grep/Glob for specific patterns, related files
- Check git log for recent changes
- Search for errors, failing tests, TODO/FIXME comments
3. **Gather additional context** as needed:
- Read local documentation files
- WebFetch for online documentation
- WebSearch for best practices, common approaches, current information
4. **Document findings** to ground questions in actual project context
**Critical Rules:**
- NEVER skip research
- Check conversation history before exploring codebase
- Questions must be grounded in actual findings, not assumptions or base knowledge
- Route Glob, Grep, WebSearch, WebFetch, and multi-file Read through `Task/Explore` — never call them directly in main context
- Include conversation-relevant context (file paths, errors, prior decisions) in every Explore prompt — Explore cannot see prior turns
For detailed research strategies, patterns, and examples, see [references/research-strategies.md](references/research-strategies.md).
### Phase 2: Generate Targeted Questions
Based on research findings, formulate 1-6 questions that will clarify the ambiguity.
**Question Guidelines:**
- **Grounded**: Every option comes from research (codebase findings, documentation, common patterns)
- **Specific**: Avoid vague options like "Other approach"
- **Multiple choice**: Provide 2-4 concrete options per question
- **Focused**: Each question addresses one decision point
- **Contextual**: Include brief explanations of trade-offs
**Number of Questions:**
- **1-2 questions**: Simple ambiguity (which file? which approach?)
- **3-4 questions**: Moderate complexity (scope + approach + validation)
- **5-6 questions**: Complex scenarios (major feature with multiple decision points)
For question templates, effective patterns, and examples, see [references/question-patterns.md](references/question-patterns.md).
### Phase 3: Get Clarification
Use the AskUserQuestion tool to present your research-grounded questions.
**AskUserQuestion Format:**
```
- question: Clear, specific question ending with ?
- header: Short label (max 12 chars) for UI display
- multiSelect: false (unless choices aren't mutually exclusive)
- options: Array of 2-4 specific choices from research
- label: Concise choice text (1-5 words)
- description: Context about this option (trade-offs, implications)
```
**Important:** Always include multiSelect field (true/false). User can always select "Other" for custom input.
### Phase 4: Execute with Context
Proceed with the original user request using:
- Original prompt intent
- Clarification answers from user
- Research findings and context
- Conversation history
Execute the request as if it had been clear from the start.
## Examples
### Example 1: Skill Invocation → Research → Questions → Execution
**Hook evaluation:** Determined prompt is vague
**Original prompt:** "fix the bug"
**Skill invoked:** Yes (prompt lacks target and context)
**Research plan:**
1. Check conversation history for recent errors
2. Explore codebase for failing tests
3. Grep for TODO/FIXME comments
4. Check git log for recent problem areas
**Research findings:**
- Recent conversation mentions login failures
- auth.py:145 has try/catch swallowing errors
- Tests failing in test_auth.py
**Questions generated:**
1. Which bug are you referring to?
- Login authentication failure (auth.py:145)
- Session timeout issues (session.py:89)
- Other
**User answer:** Login authentication failure
**Execution:** Fix the error handling in auth.py:145 that's causing login failures
### Example 2: Clear Prompt (Skill Not Invoked)
**Original prompt:** "Refactor the getUserById function in src/api/users.ts to use async/await instead of promises"
**Hook evaluation:** Passes all checks
- Specific target: getUserById in src/api/users.ts
- Clear action: refactor to async/await
- Success criteria: use async/await instead of promises
**Skill invoked:** No (prompt is clear, proceeds immediately without skill invocation)
For comprehensive examples showing various prompt types and transformations, see [references/examples.md](references/examples.md).
## Key Principles
1. **Assume Vagueness**: Skill is only invoked for vague prompts (evaluation done by hook)
2. **Research First**: Always gather context before formulating questions
3. **Ground Questions**: Use research findings, not assumptions or base knowledge
4. **Be Specific**: Provide concrete options from actual codebase/context
5. **Stay Focused**: Max 1-6 questions, each addressing one decision point
6. **Systematic Approach**: Follow 4-phase workflow (Research → Questions → Clarify → Execute)
## Progressive Disclosure
This SKILL.md contains the core workflow and essentials. For deeper guidance:
- **Research strategies**: [references/research-strategies.md](references/research-strategies.md)
- **Question patterns**: [references/question-patterns.md](references/question-patterns.md)
- **Comprehensive examples**: [references/examples.md](references/examples.md)
Load these references only when detailed guidance is needed on specific aspects of prompt improvement.
提问、分享技巧或报告无法使用。按“有帮助”数量排序。