Use this skill when an operator wants to run a repeatable coding-agent workflow through Qwen Code rather than a one-off model chat. The operator installs the Qwen Code CLI, authenticates a model provider, launches the terminal agent, then uses the runtime for codebase navigation, edits, tests, MCP-backed tools, subagents, Auto-Skills, sandboxed execution, or delegated coding work from another agent.

Invoke this instead of using the product normally when the job is to make Qwen Code part of an agent/operator workflow: bootstrapping an open-source terminal coding agent on a project, standardizing provider authentication and settings, running agentic coding sessions with MCP and skills enabled, or letting Claude, Codex, or another ACP-capable agent delegate coding tasks to Qwen Code through the upstream Qwen Code Claw reference.

The scope boundary is the operational use of Qwen Code as a coding-agent runtime. This is not a generic Qwen model, SDK, npm package, or terminal app listing; it covers the source-backed workflow for installing, authenticating, launching, configuring, and delegating to Qwen Code for repeatable software engineering tasks.

Source-backed install path from the upstream README: npm install -g @qwen-code/qwen-code@latest

Alternative upstream install path: brew install qwen-code

Quick start documented upstream: run qwen, then run /auth inside the session.

For cross-agent delegation, use the upstream Qwen Code Claw skill reference from the repository and follow its instructions to install, authenticate, and use Qwen Code through ACP-compatible delegation.