Qwen Code

# Qwen Code Skill (background-first)

Use bash background mode for non-interactive coding work with Qwen Code CLI.

The Pattern: workdir + background

```bash # Start Qwen Code in target directory bash workdir:~/project background:true yieldMs:30000 command:"qwen -p 'Build a Flask API'" # Returns sessionId for tracking

# Monitor progress process action:log sessionId:XXX

# Check if done process action:poll sessionId:XXX

# Send input (if Qwen asks a question) process action:write sessionId:XXX data:"y"

# Kill if needed process action:kill sessionId:XXX ```

Why workdir matters: Agent wakes up in a focused directory, doesn't wander off reading unrelated files.

---

Quick Start

Prerequisites

```bash # Install Qwen Code CLI npm install -g @qwen-code/qwen-code@latest

# Verify installation qwen --version

# Authenticate (Option 1: OAuth) qwen auth login

# Or Option 2: API Key export DASHSCOPE_API_KEY="sk-xxx" ```

Basic Usage

```bash # Check status scripts/qwen-code.js status

# Run a task scripts/qwen-code.js run "Create a Flask API"

# Code review scripts/qwen-code.js review src/app.ts

# Headless mode (JSON output) scripts/qwen-code.js headless "Analyze code" -o json ```

---

Commands

| Command | Description | Example | |---------|-------------|---------| | `status` | Check Qwen Code status and authentication | `scripts/qwen-code.js status` | | `run <task>` | Execute programming task | `scripts/qwen-code.js run "Create REST API"` | | `review <file>` | Code review and analysis | `scripts/qwen-code.js review src/main.py` | | `headless <task>` | Headless mode (JSON output) | `scripts/qwen-code.js headless "Analyze" -o json` | | `help` | Show help information | `scripts/qwen-code.js help` |

---

OpenClaw Integration

Background Execution

```bash # Basic task bash workdir:~/project background:true yieldMs:30000 \ command:"qwen -p 'Create Python Flask API'"

# Specify model bash workdir:~/project background:true yieldMs:30000 \ command:"qwen -p 'Analyze code structure' -m qwen3-coder-plus"

# YOLO mode (auto-approve) bash workdir:~/project background:true yieldMs:30000 \ command:"qwen -p 'Refactor this function' -y" ```

Process Management

```bash # View logs process action:log sessionId:XXX

# Check completion process action:poll sessionId:XXX

# Send input (if Qwen asks) process action:write sessionId:XXX data:"y" ```

Headless Mode (Automation/CI/CD)

```bash # JSON output qwen -p "Analyze code structure" --output-format json

# Pipeline operations git diff | qwen -p "Generate commit message"

# Batch processing find src -name "*.ts" | xargs -I {} qwen -p "Review {}" ```

---

Models

Qwen Code supports Alibaba Cloud models:

• `qwen3.5-plus` - General purpose (default)
• `qwen3-coder-plus` - Coding specialized
• `qwen3-coder-next` - Latest coding model
• `qwen3-max-2026-01-23` - Most capable

Specify model: ```bash bash workdir:~/project background:true yieldMs:30000 \ command:"qwen -p 'Refactor this' -m qwen3-coder-plus" ```

---

Authentication

OAuth (Recommended)

```bash qwen auth login ```

Opens browser for OAuth flow. Token auto-refreshes.

API Key

```bash export DASHSCOPE_API_KEY="sk-xxx" ```

Get key from: https://dashscope.console.aliyun.com/

---

⚠️ Rules

1. Respect tool choice — if user asks for Qwen, use Qwen. NEVER offer to build it yourself!
2. Be patient — don't kill sessions because they're "slow"
3. Monitor with process:log — check progress without interfering
4. YOLO mode for building — `--yolo` auto-approves changes (use in workspace only)
5. Review mode for safety — production code should use review mode
6. Parallel is OK — run many Qwen processes at once for batch work
7. NEVER start Qwen in ~/clawd/ — it'll read your soul docs! Use target project dir or /tmp
8. Workspace safety — YOLO mode is safe in `agents.defaults.workspace`, not elsewhere

---

For

• Developers using Qwen Code for programming tasks
• Teams needing code review and analysis
• Automation scripts and CI/CD integration
• OpenClaw Sub-Agent and Skills management
• Batch code analysis and refactoring

Not For

• Environments without Qwen Code CLI installed
• GUI-based interaction requirements
• Non-Alibaba Cloud LLM users
• Offline environments (requires network connection)

---

Security & Boundaries

| Component | Behavior | Executes Shell Commands? | |-----------|----------|-------------------------| | `scripts/qwen-code.js` | Wraps Qwen Code CLI commands | Yes (via `qwen` command) | | `references/qwen-cli-commands.md` | Command reference documentation | No (plain text) | | `assets/examples/` | Example code files | No (static files) |

⚠️ Security Notes

• This Skill does not execute code directly, only calls Qwen Code CLI
• All code generation and modifications require user confirmation
• Use review mode in production environments
• Disable YOLO mode for sensitive projects

---

Examples

See `assets/examples/` for complete examples:

| Example | Description | |---------|-------------| | `basic-task.example.sh` | Basic task execution | | `code-review.example.sh` | Code review workflow | | `ci-cd.example.yml` | GitHub Actions integration | | `headless-mode.example.js` | Node.js automation example |

---

References

• 📖 Qwen Code Official Docs
• 📝 Command Reference
• 📦 Example Code
• 🦌 OpenClaw Documentation

---

Troubleshooting

"qwen: command not found"

```bash npm install -g @qwen-code/qwen-code@latest ```

"Authentication required"

```bash qwen auth login # Or set API key export DASHSCOPE_API_KEY="sk-xxx" ```

Session stuck/waiting for input

```bash # Check what Qwen is asking process action:log sessionId:XXX

# Send approval process action:write sessionId:XXX data:"y" ```

Kill stuck session

```bash process action:kill sessionId:XXX ```

---

*Qwen Code Skill 🦌 - Your AI coding partner powered by Alibaba Cloud*