Content
# 🤖 dev-planner-skill
> **Conversational AI Development Planner** —— Empowers anyone to turn an idea into a complete project through conversation.
---
## ⚡ Integration Guide for AI Programming Tools
> This section is written for AIs. If you are Claude Code, Antigravity, Codex CLI, Gemini CLI, or another AI programming assistant, please read this section to learn how to load and use this Skill.
### What This Skill Is
`dev-planner` is a conversational development planning workflow, with `SKILL.md` as its core file. After loading, you will gain the following capabilities:
- Guide users to refine development requirements through multi-turn option dialogues (zero jargon, beginner-friendly)
- Automatically generate three enterprise-level development documents (`DEV_DOCUMENT.md` / `API 文档` / `风格指南`)
- Orchestrate Agent sub-teams (backend / frontend / testing / documentation) for parallel development
- Automatically detect and call available MCP tools in the environment (GitHub / database / Docker / web_search)
- Document-driven development throughout the process, node testing gating, Chinese Git commits
---
### 📥 File Download (Universal for All Platforms)
```bash
# Download the core Skill file
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md -o SKILL.md
# Download the complete template package (recommended to download together, AI will reference when generating documents)
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/dev-doc-template.md -o dev-doc-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/api-doc-template.md -o api-doc-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/style-guide-template.md -o style-guide-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/plan-template.md -o plan-template.md
# Or clone the entire repository
git clone https://github.com/cat9999aaa/dev-planner-skill.git
```
---
### 🔧 Integration Methods for Each Platform
---
#### Antigravity
Antigravity has two installation methods, choose according to the scope of use:
**Method 1: Project-Level Installation (Only Effective for the Current Project)**
```bash
# Execute in your project root directory
mkdir -p .agent/skills/dev-planner
# Download SKILL.md to the skill directory
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o .agent/skills/dev-planner/SKILL.md
# Download template files (optional, improves document generation quality)
mkdir -p .agent/skills/dev-planner/references
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/dev-doc-template.md \
-o .agent/skills/dev-planner/references/dev-doc-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/api-doc-template.md \
-o .agent/skills/dev-planner/references/api-doc-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/style-guide-template.md \
-o .agent/skills/dev-planner/references/style-guide-template.md
```
Final directory structure:
```
Your Project/
└── .agent/
└── skills/
└── dev-planner/
├── SKILL.md ← Core file
└── references/ ← Template files (optional)
```
**Method 2: Global Installation (Effective for All Projects)**
```bash
# Linux / macOS
mkdir -p ~/.antigravity/skills/dev-planner
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o ~/.antigravity/skills/dev-planner/SKILL.md
# Windows (PowerShell)
New-Item -ItemType Directory -Force "$env:USERPROFILE\.antigravity\skills\dev-planner"
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md" `
-OutFile "$env:USERPROFILE\.antigravity\skills\dev-planner\SKILL.md"
```
Alternatively, if you have already cloned the repository locally, you can use the CLI tool to link:
```bash
# Enter the skill directory
cd dev-planner-skill/dev-planner
# Link as a global skill
ag skill link --global
```
---
#### Claude Code
Claude Code loads context instructions through the `CLAUDE.md` file.
**Project-Level Installation (Recommended):**
```bash
# Download SKILL.md to the project root directory, rename it to CLAUDE.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o CLAUDE.md
# Download template files to the references/ directory at the same time
mkdir -p references
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/dev-doc-template.md \
-o references/dev-doc-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/api-doc-template.md \
-o references/api-doc-template.md
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/style-guide-template.md \
-o references/style-guide-template.md
```
**Global Installation (Effective for All Projects):**
```bash
# Linux / macOS
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o ~/.claude/CLAUDE.md
# Windows (PowerShell)
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md" `
-OutFile "$env:USERPROFILE\.claude\CLAUDE.md"
```
Claude Code automatically reads `CLAUDE.md` after startup, and automatically triggers this Skill when the user mentions keywords such as "help me develop a project".
---
#### OpenAI Codex CLI
Codex CLI loads instructions through the `codex.md` file in the project root directory.
```bash
# Execute in the project root directory
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o codex.md
# Download template files
mkdir -p references
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/references/dev-doc-template.md \
-o references/dev-doc-template.md
# (The remaining templates are the same as above, download as needed)
```
Codex CLI automatically reads `codex.md` in the current directory when it starts, without additional configuration.
---
#### Google Gemini CLI
Gemini CLI loads instructions through the `GEMINI.md` file in the project root directory (similar to Claude Code's `CLAUDE.md`).
**Project-Level Installation:**
```bash
# Execute in the project root directory
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o GEMINI.md
```
**Global Installation:**
```bash
# Linux / macOS
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o ~/.gemini/GEMINI.md
```
---
#### Cursor
Cursor loads project-level rules through the `.cursorrules` file (or the new `.cursor/rules/` directory).
```bash
# Method 1: Use .cursorrules (old version compatibility)
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o .cursorrules
# Method 2: Use the new version rule directory (Cursor 0.43+)
mkdir -p .cursor/rules
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o .cursor/rules/dev-planner.mdc
```
---
#### Aider
Aider loads custom instructions through the `--system` parameter or configuration file.
```bash
# Download SKILL.md first
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o .aider-skill.md
# Inject at startup
aider --system "$(cat .aider-skill.md)"
```
Or write to the Aider configuration file `.aider.conf.yml`:
```yaml
# .aider.conf.yml
system: |
# Paste the complete content of SKILL.md here
```
---
#### Continue.dev (VS Code / JetBrains Plugin)
Add a system prompt in the Continue configuration file `~/.continue/config.json`:
```bash
# Download SKILL.md content first
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md \
-o /tmp/dev-planner-skill.md
```
Then edit `~/.continue/config.json`:
```json
{
"models": [
{
"title": "Claude with Dev Planner",
"provider": "anthropic",
"model": "claude-opus-4-6",
"systemMessage": "<Paste the complete content of SKILL.md>"
}
]
}
```
---
#### Universal Solution (Applicable to Any AI Tool That Supports System Prompts)
If the tool you are using is not in the list above, as long as it supports custom system prompts, you can use this method:
```bash
# Get the content of SKILL.md and copy it to the clipboard
curl -L https://raw.githubusercontent.com/cat9999aaa/dev-planner-skill/main/SKILL.md
```
Paste the output content into the "System Prompt", "Custom Instructions" or "Role Setting" area of your tool.
---
### 📁 File Structure Description
```
dev-planner/
├── SKILL.md # Core Skill file, load this file to activate all capabilities
└── references/
├── dev-doc-template.md # DEV_DOCUMENT.md Super detailed template (including architecture diagram/ER diagram/flowchart)
├── api-doc-template.md # API interface documentation template (including request response examples/error codes)
├── style-guide-template.md # Code specification + design system template
└── plan-template.md # Planning scheme output template
```
**Minimal Integration**: Only `SKILL.md` is required to run. The templates in `references/` are used to improve document quality and are recommended to be downloaded together.
---
### 🚦 Trigger Keywords
Any of the following words will trigger the dev-planner Skill:
```
"帮我开发..." "我想做一个..." "我有个想法..."
"做个网站" "做个APP" "做个脚本"
"做个机器人" "做个系统" "plan模式"
"对话式开发" "生成开发文档" "帮我规划项目"
```
Even if the description is very vague (such as "I want to make something"), this Skill should be triggered to enter the dialogue planning mode.
---
### 🔄 Workflow Overview (For AI Understanding)
```
User describes idea
│
▼
Phase 1 PLAN MODE
8-step option dialogue → Collect: Project type/scale/technology stack/function/UI/deployment
│ User confirms plan
▼
Phase 2 DOC GEN
Generate three documents: DEV_DOCUMENT.md + API documentation + style guide
│ User confirms document
▼
Phase 3 MCP SETUP
Detect available tools: GitHub MCP / Database MCP / Docker MCP / web_search
│
▼
Phase 4 TEAM DEV
Orchestrate Agent team for parallel development:
Backend Agent + Frontend Agent + Test Agent + Documentation Agent
Each node: Development → MCP assistance → Testing → Submission → Documentation synchronization
│ All nodes completed
▼
Phase 5 DELIVERY
Final report + technical manual + user manual + step-by-step deployment tutorial
```
---
### ⚠️ Important Constraints (AI Must Obey)
1. **Document Priority**: All development strictly follows the generated three documents, and the technology stack must not be changed without authorization
2. **Test Gating**: If the test of each node fails, it must not be submitted and must not be skipped
3. **MCP Priority**: When MCP tools are available, call them first, do not manually simulate
4. **Chinese Submission**: All Git commit messages must be in Chinese
5. **Document Synchronization**: Immediately update the three documents after each node is completed
---
---
## 👋 Introduction for Humans
### What This Project Is
dev-planner-skill is an **AI programming Skill (skill package)**, specifically designed to solve a problem:
> **I have an idea, but I don’t know how to develop it, and I don’t know where to start.**
The AI with this Skill installed will become a patient development team leader. It won't ask you to fill in the technology stack or write requirements documents right away—it will ask you "What do you want to do" like a friend, and then turn the vague idea in your mind into a clear development plan step by step through a series of multiple-choice questions.
After the plan is confirmed, the AI will automatically generate three enterprise-level documents, and then summon a virtual Agent development team (backend/frontend/testing/documentation) to develop the project in parallel, without you needing to understand any technology throughout the process.
---
### Why Do This
This project stems from a real observation:
AI programming tools are getting stronger and stronger, but **the threshold is getting higher and higher**. You have to know how to write Prompts, you have to know what technology stack to choose, and you have to understand what an API is... For real "newbies", these tools seem to have nothing to do with them.
Another problem is that even for experienced developers, the process of developing with AI is often chaotic: AI changes here today, forgets there tomorrow, the code becomes more and more chaotic, there is never any documentation, and in the end, they don’t even know what the project is. status.
dev-planner wants to solve these two problems:
1. **Make it usable for newbies**: Multiple-choice questions throughout the conversation, no need to understand any technical terms
2. **Make the development process orderly**: Document-driven, test-gated, every step is recorded, and the project status is clear at any time
---
### How to Use
#### If You Are Using Claude Code
```bash
# Install
claude skill install https://github.com/cat9999aaa/dev-planner-skill/releases/latest/download/dev-planner.skill
# Then just talk, for example:
# "I want to make an accounting App"
# "Help me develop a script to automatically post Weibo"
# "I want to make an internal leave approval system for the company"
```
#### If You Are Using Other AI Tools
1. Open https://github.com/cat9999aaa/dev-planner-skill
2. Download `dev-planner/SKILL.md`
3. Copy and paste the content into the "Custom System Prompt" area of the AI tool you are using
4. Start the conversation
#### Example Conversation
```
You: I want to create a website to manage my reading notes.
AI: 👋 Hello! Based on your description, this project belongs to:
🌐 A. Website/Web Application
📱 B. Mobile APP
...
(You choose A)
AI: How big is this project approximately?
🌱 A. A small tool, just for my own use
...
(You choose A)
... After several rounds of selection ...
AI: Planning completed! A total of 12 development nodes, expected to use Vue3 + Node.js + SQLite.
Reply "confirm" to start generating development documents.
You: confirm
AI: Three documents have been generated. Reply "start development" to launch the Agent team.
You: start development
AI: [Backend Agent] Database design completed ✅
[Frontend Agent] Project scaffolding completed ✅
[Backend Agent] Note CRUD interface completed ✅
... (Fully automatic, until completion)
```
---
### Document Structure
| File | Description |
| ------------------------------------------------ | ----------------------------------------- |
| `dev-planner/SKILL.md` | Core Skill, defines the complete workflow and specifications |
| `dev-planner/references/dev-doc-template.md` | Development document template (including architecture diagram/ER diagram/environment configuration, etc.) |
| `dev-planner/references/api-doc-template.md` | API interface document template |
| `dev-planner/references/style-guide-template.md` | Code specification + design system template |
| `dev-planner/references/plan-template.md` | Planning scheme output template |
| `dev-planner.skill` | Packaged Skill file (can be installed directly) |
---
### Contribution & Feedback
If you have ideas, find problems, or want to support more platforms, you are welcome to:
- Submit an [Issue](https://github.com/cat9999aaa/dev-planner-skill/issues)
- Submit a Pull Request
- Or directly describe in the Issue what you did with this Skill and where it is not enough
---
### License
MIT License — Use it as you like, modify it as you like, and remember to come back and share it after you have modified it well.
Connection Info
You Might Also Like
everything-claude-code
Complete Claude Code configuration collection - agents, skills, hooks,...
markitdown
MarkItDown-MCP is a lightweight server for converting URIs to Markdown.
servers
Model Context Protocol Servers
gibber-mcp
gibber-mcp is an Express.js server for cryptographic tools using MCP.
gibber-mcp
Tiny MCP server with cryptography tools, sufficient to establish end-to-end...
sora-mcp
Sora MCP Server generates and remixes videos using OpenAI's Sora 2 API.