🧠 DeepSeek V4 AI Agent Skill

An open agent skill that brings DeepSeek's power to 15+ AI coding platforms. Uses DeepSeek V3.2 now — seamless, zero-config upgrade to V4 when it launches.

Why This Skill?

DeepSeek V3.2 is one of the most capable open-weight LLMs for coding, reasoning, and analysis. This skill makes it instantly available across every major AI coding platform — Claude Code, Cursor, GitHub Copilot, Codex CLI, Gemini CLI, Windsurf, and more.

When DeepSeek V4 launches, this skill upgrades automatically. No config changes, no reinstall, no downtime. Just better results.

Key value:

One skill, 15+ platforms — install once, use everywhere
OpenAI-compatible API — works with any tool that speaks OpenAI format
Future-proof — automatic V3.2 → V4 upgrade path
Cost-effective — $0.26/M input tokens, fraction of GPT-4 pricing
No rate limits — powered by Atlas Cloud's infrastructure

Quick Start

Install via npx (Recommended)

npx skills add thoughtincode/deepseek-v4-skill

Or clone manually

git clone https://github.com/thoughtincode/deepseek-v4-skill.git
cd deepseek-v4-skill

Set your API key

export ATLAS_API_KEY="your-api-key-here"

Get your API key at atlascloud.ai — 25% bonus on first top-up

Supported Platforms

This skill works with 15+ AI coding platforms out of the box:

Platform	Type	Status
Claude Code	CLI Agent	✅ Full support
Cursor	IDE Agent	✅ Full support
GitHub Copilot	IDE Extension	✅ Full support
OpenAI Codex CLI	CLI Agent	✅ Full support
Gemini CLI	CLI Agent	✅ Full support
Windsurf	IDE Agent	✅ Full support
Kiro	IDE Agent	✅ Full support
OpenCode	CLI Agent	✅ Full support
Cline	IDE Extension	✅ Full support
Aider	CLI Agent	✅ Full support
Continue	IDE Extension	✅ Full support
Roo Code	IDE Extension	✅ Full support
AugmentCode	IDE Extension	✅ Full support
Amazon Q Developer	IDE Agent	✅ Full support
Tabnine	IDE Extension	✅ Full support
Custom Agents	Any OpenAI-compatible	✅ Full support

Platform-Specific Setup

Claude Code

The CLAUDE.md file is automatically detected. Just add the skill to your project:

npx skills add thoughtincode/deepseek-v4-skill
export ATLAS_API_KEY="your-key"

Claude Code will use the skill when you ask it to run DeepSeek tasks.

Cursor

Add to your .cursor/rules or project settings:

Use DeepSeek V4 skill for code analysis and generation tasks.
API: POST https://api.atlascloud.ai/v1/chat/completions
Model: deepseek/deepseek-v3.2

GitHub Copilot / Codex CLI

The skill's OpenAI-compatible endpoint works directly:

export OPENAI_API_BASE=https://api.atlascloud.ai/v1
export OPENAI_API_KEY=$ATLAS_API_KEY

Gemini CLI / Windsurf / Kiro

Reference the SKILL.md file in your agent configuration. The OpenAI-compatible format means zero adapter code needed.

Aider

aider --openai-api-base https://api.atlascloud.ai/v1 \
      --openai-api-key $ATLAS_API_KEY \
      --model deepseek/deepseek-v3.2

Any OpenAI-Compatible Tool

Base URL: https://api.atlascloud.ai/v1
Model:    deepseek/deepseek-v3.2
Auth:     Bearer $ATLAS_API_KEY

Features

🖥️ Code Generation & Analysis

Generate production-ready code in 50+ languages
Explain complex algorithms and data structures
Refactor and optimize existing code
Convert code between languages

🧪 Code Review & Quality

Automated PR review with actionable feedback
Security vulnerability detection
Performance bottleneck identification
Best practices enforcement

🧠 Advanced Reasoning

Multi-step problem solving
Architecture design decisions
Algorithm complexity analysis
System design and trade-off evaluation

📄 Documentation & Summarization

Generate API documentation
Summarize large codebases
Create technical specifications
Write inline comments and docstrings

🔮 Coming with V4

1M+ token context window — analyze entire repositories at once
Enhanced coding benchmarks — expected top-tier performance
Faster inference — reduced latency for real-time coding
Improved instruction following — better at complex multi-step tasks

CLI Usage

The skill includes a standalone CLI for direct interaction:

Basic Chat

npx deepseek-v4 --prompt "Explain the difference between TCP and UDP"

Code Generation

npx deepseek-v4 --prompt "Write a Redis connection pool in Go with retry logic" \
                 --system "You are an expert Go developer. Write production-ready code."

Code Review

npx deepseek-v4 --prompt "Review this code for bugs and improvements: $(cat src/main.ts)" \
                 --system "You are a senior code reviewer. Focus on bugs, security, and performance."

Summarize

npx deepseek-v4 --prompt "Summarize this codebase architecture: $(find src -name '*.ts' -exec cat {} \;)" \
                 --max-tokens 2000

CLI Options

Option	Default	Description
`--prompt`	(required)	The prompt to send
`--model`	`deepseek/deepseek-v3.2`	Model identifier
`--temperature`	`0.7`	Sampling temperature (0-2)
`--max-tokens`	`4096`	Maximum response tokens
`--system`	(none)	System prompt for role/behavior

Environment Variables

Variable	Required	Description
`ATLAS_API_KEY`	Yes	Your Atlas Cloud API key
`ATLAS_BASE_URL`	No	Custom API base URL (default: `https://api.atlascloud.ai/v1`)

Agent Skill Integration

How It Works

This skill follows the open agent skill pattern — a standardized way to add capabilities to AI coding agents. The skill provides:

SKILL.md — Machine-readable skill description with trigger conditions and API details
CLAUDE.md — Claude Code-specific configuration
src/cli.ts — Standalone CLI for direct usage and agent tool-calling
package.json — Standard npm metadata for discoverability

Cross-Platform Skill Resolution

When an agent encounters a task like "analyze this code with DeepSeek", the skill resolution works:

User Request
    ↓
Agent reads SKILL.md / CLAUDE.md
    ↓
Matches trigger: "code analysis", "DeepSeek", "review"
    ↓
Calls Atlas Cloud API (OpenAI-compatible)
    ↓
Returns DeepSeek V3.2 response
    ↓
(Seamless V4 upgrade when available)

Embedding in Your Agent

// 在任何OpenAI兼容的客户端中使用
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ATLAS_API_KEY,
  baseURL: "https://api.atlascloud.ai/v1",
});

const response = await client.chat.completions.create({
  model: "deepseek/deepseek-v3.2",
  messages: [
    { role: "system", content: "You are a senior software engineer." },
    { role: "user", content: "Review this pull request for security issues..." },
  ],
  temperature: 0.7,
  max_tokens: 4096,
});

console.log(response.choices[0].message.content);

API Reference

Endpoint

POST https://api.atlascloud.ai/v1/chat/completions

Authentication

Authorization: Bearer YOUR_ATLAS_API_KEY
Content-Type: application/json

Request Body

{
  "model": "deepseek/deepseek-v3.2",
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful coding assistant."
    },
    {
      "role": "user",
      "content": "Write a binary search implementation in Rust"
    }
  ],
  "temperature": 0.7,
  "max_tokens": 4096,
  "stream": true
}

Response Format

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1710000000,
  "model": "deepseek/deepseek-v3.2",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here's a binary search implementation in Rust:\n\n```rust\nfn binary_search<T: Ord>(arr: &[T], target: &T) -> Option<usize> {\n    let mut low = 0;\n    let mut high = arr.len();\n    while low < high {\n        let mid = low + (high - low) / 2;\n        match arr[mid].cmp(target) {\n            std::cmp::Ordering::Equal => return Some(mid),\n            std::cmp::Ordering::Less => low = mid + 1,\n            std::cmp::Ordering::Greater => high = mid,\n        }\n    }\n    None\n}\n```"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 150,
    "total_tokens": 175
  }
}

cURL Example

curl -X POST https://api.atlascloud.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ATLAS_API_KEY" \
  -d '{
    "model": "deepseek/deepseek-v3.2",
    "messages": [{"role": "user", "content": "Write a quicksort in Python"}],
    "temperature": 0.7,
    "max_tokens": 4096
  }'

Streaming Example

curl -X POST https://api.atlascloud.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ATLAS_API_KEY" \
  -d '{
    "model": "deepseek/deepseek-v3.2",
    "messages": [{"role": "user", "content": "Explain monads in Haskell"}],
    "temperature": 0.7,
    "stream": true
  }'

Python Example

import openai

client = openai.OpenAI(
    api_key="your-atlas-api-key",
    base_url="https://api.atlascloud.ai/v1",
)

response = client.chat.completions.create(
    model="deepseek/deepseek-v3.2",
    messages=[
        {"role": "system", "content": "You are a Python expert."},
        {"role": "user", "content": "Write an async web scraper with aiohttp"},
    ],
    temperature=0.7,
    max_tokens=4096,
)

print(response.choices[0].message.content)

Node.js Example

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ATLAS_API_KEY,
  baseURL: "https://api.atlascloud.ai/v1",
});

async function generateCode(prompt: string): Promise<string> {
  const response = await client.chat.completions.create({
    model: "deepseek/deepseek-v3.2",
    messages: [{ role: "user", content: prompt }],
    temperature: 0.7,
    max_tokens: 4096,
  });

  return response.choices[0].message.content ?? "";
}

const code = await generateCode("Write a REST API with Express and TypeScript");
console.log(code);

V3.2 Current vs V4 Expected

Feature	DeepSeek V3.2 (Current)	DeepSeek V4 (Expected)
Context Window	128K tokens	1M+ tokens
Coding (HumanEval)	90.2%	Expected 95%+
Reasoning (MATH)	90.0%	Expected 95%+
Multilingual	50+ languages	100+ languages
Architecture	MoE	Enhanced MoE
Latency	Fast	Faster
Price (Input)	$0.26/M tokens	TBD
Price (Output)	$0.38/M tokens	TBD
API Compatibility	OpenAI-compatible	OpenAI-compatible
Upgrade Path	—	Automatic via this skill

What Happens When V4 Launches?

Atlas Cloud adds DeepSeek V4 to their API
This skill's model identifier updates to deepseek/deepseek-v4
Your existing integration keeps working — no code changes needed
You get V4's improvements immediately

Zero downtime. Zero config changes. Just better results.

Pricing

DeepSeek V3.2 (Current)

Metric	Price
Input tokens	$0.26 / million tokens
Output tokens	$0.38 / million tokens
Context window	128K tokens
Rate limits	None (Atlas Cloud)

Cost Comparison

Model	Input ($/M)	Output ($/M)	Savings vs GPT-4
DeepSeek V3.2	$0.26	$0.38	~95% cheaper
GPT-4o	$2.50	$10.00	—
Claude 3.5 Sonnet	$3.00	$15.00	—
Gemini 1.5 Pro	$3.50	$10.50	—

Example Cost Estimates

Use Case	Tokens/Day	Daily Cost	Monthly Cost
Individual developer	~100K	~$0.03	~$1
Small team (5 devs)	~500K	~$0.15	~$5
CI/CD code review	~1M	~$0.32	~$10
Enterprise (50 devs)	~10M	~$3.20	~$100

💡 Get 25% bonus on your first top-up at atlascloud.ai

Use Cases

1. Automated Code Review in CI/CD

# .github/workflows/deepseek-review.yml
name: DeepSeek Code Review
on: [pull_request]
jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Review with DeepSeek
        run: |
          DIFF=$(git diff origin/main...HEAD)
          npx deepseek-v4 \
            --prompt "Review this PR diff for bugs, security issues, and improvements:\n$DIFF" \
            --system "You are a senior code reviewer. Be thorough but constructive." \
            --max-tokens 4096
        env:
          ATLAS_API_KEY: ${{ secrets.ATLAS_API_KEY }}

2. Code Generation Pipeline

# 生成完整的CRUD API
npx deepseek-v4 \
  --prompt "Generate a complete CRUD REST API for a blog platform with posts, comments, and users. Use Express.js, TypeScript, and Prisma ORM." \
  --system "Generate production-ready code with error handling, validation, and types." \
  --max-tokens 8192

3. Architecture Documentation

# 从代码生成架构文档
npx deepseek-v4 \
  --prompt "Analyze this codebase and generate architecture documentation: $(find src -name '*.ts' | head -20 | xargs cat)" \
  --system "You are a technical writer. Create clear, structured architecture docs." \
  --max-tokens 4096

4. Bug Investigation

# 分析错误日志
npx deepseek-v4 \
  --prompt "Investigate this error and suggest fixes: $(cat error.log | tail -50)" \
  --system "You are a debugging expert. Identify root cause and provide fix."

Configuration

Environment Variables

Variable	Required	Default	Description
`ATLAS_API_KEY`	Yes	—	Atlas Cloud API key
`ATLAS_BASE_URL`	No	`https://api.atlascloud.ai/v1`	API base URL
`DEEPSEEK_MODEL`	No	`deepseek/deepseek-v3.2`	Model identifier
`DEEPSEEK_TEMPERATURE`	No	`0.7`	Default temperature
`DEEPSEEK_MAX_TOKENS`	No	`4096`	Default max tokens

Custom Configuration File

Create a .deepseek-v4.json in your project root:

{
  "model": "deepseek/deepseek-v3.2",
  "temperature": 0.7,
  "maxTokens": 4096,
  "systemPrompt": "You are a senior software engineer specializing in TypeScript and React.",
  "baseUrl": "https://api.atlascloud.ai/v1"
}

Troubleshooting

Common Issues

401 Unauthorized

Your API key is missing or invalid.

# 检查API密钥是否设置
echo $ATLAS_API_KEY

# 重新设置
export ATLAS_API_KEY="your-key-here"

Get your key at atlascloud.ai.

Model not found

Ensure you're using the correct model identifier:

deepseek/deepseek-v3.2    ✅ Correct
deepseek-v3.2              ❌ Missing prefix
deepseek-v4                ❌ Not yet available

Timeout errors

For large prompts, increase the timeout:

# 使用较小的max-tokens值
npx deepseek-v4 --prompt "..." --max-tokens 2000

Streaming not working

Streaming is enabled by default. If your environment doesn't support it, the CLI falls back to non-streaming mode automatically.

Contributing

Contributions are welcome! Please read our guidelines:

Fork the repository
Create your feature branch: git checkout -b feature/my-feature
Commit your changes: git commit -m "Add my feature"
Push to the branch: git push origin feature/my-feature
Open a Pull Request

Development

git clone https://github.com/thoughtincode/deepseek-v4-skill.git
cd deepseek-v4-skill
npm install
npm run dev

Trust & Reliability

Aspect	Detail
API Provider	Atlas Cloud — enterprise-grade AI infrastructure
Uptime	99.9% SLA
Data Privacy	No training on your data
Compliance	SOC2, GDPR compliant
Support	24/7 technical support
Rate Limits	None — scale without limits
Billing	Pay-as-you-go, no minimums

Links

Atlas Cloud: atlascloud.ai — Get your API key (25% first top-up bonus)
DeepSeek: deepseek.com — Model documentation
GitHub: thoughtincode/deepseek-v4-skill
Issues: Report a bug

License

MIT License — see LICENSE for details.

Get Started Now

# 1. 获取API密钥
# 访问 https://www.atlascloud.ai?ref=JPM683&utm_source=github&utm_campaign=deepseek-v4-skill（首充25%奖励）

# 2. 设置密钥
export ATLAS_API_KEY="your-key"

# 3. 安装技能
npx skills add thoughtincode/deepseek-v4-skill

# 4. 开始使用
npx deepseek-v4 --prompt "Write a binary search tree in Rust"

DeepSeek V3.2 today. V4 tomorrow. Zero changes needed.

Built for developers who want the best AI coding assistant at a fraction of the cost.
Get your Atlas Cloud API key · Report Bug · Request Feature

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
src		src
.env.example		.env.example
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
SKILL.md		SKILL.md
package.json		package.json

Folders and files

Latest commit

History

Repository files navigation

🧠 DeepSeek V4 AI Agent Skill

Why This Skill?

Quick Start

Install via npx (Recommended)

Or clone manually

Set your API key

Supported Platforms

Platform-Specific Setup

Features

🖥️ Code Generation & Analysis

🧪 Code Review & Quality

🧠 Advanced Reasoning

📄 Documentation & Summarization

🔮 Coming with V4

CLI Usage

Basic Chat

Code Generation

Code Review

Summarize

CLI Options

Environment Variables

Agent Skill Integration

How It Works

Cross-Platform Skill Resolution

Embedding in Your Agent

API Reference

Endpoint

Authentication

Request Body

Response Format

cURL Example

Streaming Example

Python Example

Node.js Example

V3.2 Current vs V4 Expected

What Happens When V4 Launches?

Pricing

DeepSeek V3.2 (Current)

Cost Comparison

Example Cost Estimates

Use Cases

1. Automated Code Review in CI/CD

2. Code Generation Pipeline

3. Architecture Documentation

4. Bug Investigation

Configuration

Environment Variables

Custom Configuration File

Troubleshooting

Common Issues

Contributing

Development

Trust & Reliability

Links

License

Get Started Now

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages